4 Results

4.1 Data collection and sources

The data presented in this thesis span a period from 2005-2017 and were collected and analysed over a four year period from 2013-2017. Consequently, the papers that form this thesis were written and published at different times. For this reason, the study period and the populations described in each paper varies slightly. However, they differ only marginally from the final data summary described below. The results of papers I-VI are summarized in their respective sub-chapters.

When data from all registries were taken together, individual-level information was available for 375,383 Icelandic citizens, of which 183,544 were female and 181,316 were male. Gender was not registered for 10,523 individuals. The exact date of birth was available for 366,188, and birth-year for the rest. The median birth-year for the whole study population was 1979 (IQR 1958-1997). Death was registered for 12,308 individuals.

The study often examined data stratified by birth-cohort. The number of children in each birth-cohort who contributed data to the present study is shown in Table 4.1.

Table 4.1: Demographic information regarding birth-cohorts included in the study. The number of children and proportion who are male is presented. The number of children in each cohort who had registered immigration or emigration from Iceland before four years of age is shown.

Birth-cohort	No. children	Proportion male (%)	No. moved
2005	4,803	51.5	578
2006	4,887	51.4	572
2007	4,993	51.6	567
2008	5,153	51.7	571
2009	5,331	51.7	553
2010	5,203	51.4	525
2011	4,849	51.7	473
2012	4,841	51.2	430
2013	4,566	49.4	344
2014	4,527	51.1	223
2015	4,198	51.3	144
2016	4,112	50.7	26

4.1.1 Statistics Iceland

Statistics Iceland (www.statice.is) provided data on the immigration and emigration of all Icelandic children zero to four years of age from 2005-2017. Of the 57,695 Icelandic children born 2005 or later, 5,577 moved to or from the country 6,847 times. The proportion of children in each birth-cohort who moved at least once before five years of age, was consistently 9%-12% of those birth-cohorts who had full five year follow-up time with regards to immigration and emigration (birth-cohorts 2005-2012).

4.1.2 Landspitali University Hospital patient registry

All visits and hospitalizations with ICD-10 diagnostic codes compatible with respiratory infections (Table 3.1), and procedural codes compatible with tympanostomy tube procedures (Table 3.2), were extracted from Landspitali’s patient registry. The number of visits and hospitalizations of all age-groups corresponding to each of the study’s ICD-10 codes recorded as the primary diagnosis, are shown in Table 4.2.

Table 4.2: The number of visits and hospitalizations with International Classification of Diseases, 10th revision (ICD-10) codes listed as the primary diagnosis. Hospital visits and hospitalizations are extracted from Landspitali University Hospital’s patient registry between 1 January 2005 and 31 December 2017. Primary care visits are obtained from the Primary Care Registry of the Icelandic Directorate of Health from 1 January 2005 to 31 December 2015.

ICD-10 code	Disease	Hospital visits	Hospitalizations	Primary care visits
A40	Streptococcal sepsis	37	135	68
A41	Other sepsis	370	777	279
A48	Other bacterial diseases, not elsewhere classified	5	28	10
A49	Bacterial infection of unspecified site	123	26	1,861
B00	Herpesviral [herpes simplex] infections	497	22	2,176
B08	Other viral infections characterized by skin and mucous membrane lesions, not elsewhere classified	76	1	655
B33	Other viral diseases, not elsewhere classified	32	4	106
B34	Viral infection of unspecified site	25,601	528	329,179
B95	Streptococcus, Staphylococcus, and Enterococcus as the cause of diseases classified elsewhere	12	4	40
B96	Other bacterial agents as the cause of diseases classified elsewhere	5	7	29
G00	Bacterial meningitis,not elsewhere classified	79	60	3
H65	Nonsuppurative otitis media	2,803	75	38,585
H66	Suppurative and unspecified otitis media	11,647	244	160,086
H70	Mastoiditis and related conditions	164	86	259
H72	Perforation of tympanic membrane	1,270	233	1,947
H73	Other disorders of tympanic membrane	67	3	727
J00	Acute nasopharyngitis [common cold]	3,525	49	124,984
J01	Acute sinusitis	4,625	113	152,076
J02	Acute pharyngitis	1,869	44	124,874
J03	Acute tonsillitis	5,019	213	106,491
J04	Acute laryngitis and tracheitis	983	38	19,288
J05	Acute obstructive laryngitis [croup] and epiglottitis	2,738	40	3,148
J06	Acute upper respiratory infections of multiple and unspecified sites	3,649	94	110,236
J09	Influenza due to certain identified influenza viruses	250	185	9
J10	Influenza due to other identified influenza virus	282	151	699
J11	Influenza due to unidentified influenza virus	1,003	77	34,949
J12	Viral pneumonia, not elsewhere classified	206	189	189
J13	Pneumonia due to Streptococcus pneumoniae	129	265	80
J14	Pneumonia due to Hemophilus influenzae	18	44	34
J15	Bacterial pneumonia, not elsewhere classified	2,489	1,129	1,870
J16	Pneumonia due to other infectious organisms, not elsewhere classified	60	37	62
J17	Pneumonia in diseases classified elsewhere	17	15	38
J18	Pneumonia, unspecified organism	8,576	4,501	66,232
J20	Acute bronchitis	2,431	297	148,963
J21	Acute bronchiolitis	2,874	707	6,178
J22	Unspecified acute lower respiratory infection	356	55	9,425
J32	Chronic sinusitis	3,298	405	52,899
J36	Peritonsillar abscess	1,095	254	1,239
J40	Bronchitis, not specified as acute or chronic	893	49	77,272
J85	Abscess of lung and mediastinum	98	41	24
J86	Pyothorax	20	62	48
J90	Pleural effusion, not elsewhere classified	560	409	599
N30	Cystitis	6,112	568	133,560
N39	Other disorders of urinary system	12,901	2,868	36,154
R05	Cough	2,471	11	83,948
R50	Fever of other and unknown origin	3,433	557	27,121

In total, 169,585 contacts (of 74,740 individuals) with the study’s ICD-10 codes were recorded, of which 135,841 (64,090) were visits to outpatient clinics or emergency departments and 33,744 (20,318) were hospital admissions. The highest recorded number of visits of a single individual was 170 and the most admissions was 31. The number of procedures performed at Landspitali University Hospital is shown in Table 4.3.

Table 4.3: The number of NOMESCO Classification of Surgical Procedures (NCSP) performed at Landspitali University Hospital between 1 January 2005 and 31 December 2017. The data is presented for all age-groups

NCSP code	Description	Number of procedures
EMSB00	Excision of lesion of tonsil or adenoid	1
EMSB10	Tonsillectomy	88
EMSB15	Intracapsular destruction of tonsils	2
EMSB20	Adenotonsillectomy	101
EMSB30	Adenotomy	170
EMSB99	Other excision on tonsils and adenoids	2
EMSW99	Other operation on tonsil or adenoids	1
DCSA10	Paracentesis of tympanic membrane	289
DCSA20	Insertion of ventilating tube through tympanic membrane	340
DCSW00	Removal of ventilating tube from tympanic membrane	0

The age distribution of visits and hospital admissions is shown in Figure 4.1. Though children and young adults comprised most of the visits due to study diagnoses, older adults made up the largest number of hospitalizations.

Figure 4.1: The total number of contacts to Landspitali University Hospital and Primary Care Centers in Iceland associated with International Classification of Diseases, 10th revision (ICD-10) codes compatible with respiratory infections. Data on hospital contacts were extracted from Landspitali University Hospital’s patient registry from 1 January 2005 to 31 December 2017. Primary care contacts were obtained from the Primary Care Registry of the Icelandic Directorate of Health from 1 January 2005 to 31 December 2015. The number of contacts are shown as a function of age. Different Y-axis scales are used for each category. The figure demonstrates disproportionate hospitalizations among older adults, compared to primary care contacts and hospital visits.

In addition to the increased frequency of hospitalization among older adults, the cost associated with each visit and hospitalization was higher (Figure 4.2).

Figure 4.2: The distribution in the cost associated with contacts to Landspitali University Hospital was extracted from the patient registry for the period of 1 January 2005 to 31 December 2017. Costs are presented in Icelandic Krona (ISK) at the value of the calendar year in which they occurred. Distributions are presented separately for visits and hospitalizations, and are further divided into age-groups. The X-axis has been logarithmically scaled. The figure shows that costs associated with visits range from 1,000 ISK to 100,000 ISK, while hospitalizations range from 100,000 ISK to 10,000,000 ISK. As age increases, the distribution shifts towards higher costs.

4.1.3 The Primary Care Registry

The Primary Care Registry contains information on all primary care health contacts for the period 2005-2015. All contacts associated with the ICD-10 codes listed in Table 3.1, regardless of age, were extracted. A total of 1,963,439 separate contacts were recorded between 298,307 individual patients and 1,266 different physicians. The most visits recorded for a single individual was 212. The distribution of contacts by age can be seen in Figure 4.1.

4.1.4 The National Vaccine Registry

The National Vaccine Registry contains information on all administered vaccine doses for the period 2005-2017. All recorded pneumococcal vaccine doses were extracted using ATC code J07 and sub-levels. A total of 110,712 doses of pneumococcal vaccines were administered to 51,601 individuals during the study period. The monthly number of administered doses per age-group and vaccine is shown in Figure 4.3.

Figure 4.3: The monthly number of administered doses of pneumococcal vaccines in Iceland from 1 January 2005 to 31 December 2017. Data were extracted from the National Vaccine Registry of the Icelandic Directorate of Health using Anatomical Therapeutic Chemical code J07. Records were available for four different pneumococcal vaccines; the 23-valent pneumococcal polysaccharide vaccine (23-PPV, purple), seven-valent pneumococcal conjugate vaccine (PCV7, red), the 10-valent pneumococcal conjugate vaccine (PCV10, blue) and the 13-valent pneumococcal conjugate vaccine (PCV13, green). The introduction of PCV10 into the Icelandic pediatric vaccination program on 1 April 2011, is shown with a dotted line. Different Y-axis scales are used for each category of age. The figure demonstrates the abrupt and sustained increase in the number of administered doses of PCV10 following introduction into the vaccine program. Adult vaccination with 23-PPV does not appear to have increased during the study period.

Table 4.4 shows the number of children in each birth-cohort who had received zero, one, two, or three doses of a pneumococcal conjugate vaccine by four years of age. An abrupt shift is observed in the first vaccine eligible cohort, of which over 90% received three or more doses of a pneumococcal conjugate vaccine. Children who moved to or from the country before four years of age, were excluded from the table.

Table 4.4: The number of children in each birth-cohort who received zero to three doses of a pneumococcal conjugate vaccine before four years of age are shown, with the percentage given within parentheses. Children who were documented to have immigrated or emigrated from Iceland before four years of where excluded. For the remaining children, data were obtained from the National Vaccine Registry of the Icelandic Directorate of Health using Anatomical Therapeutic Chemical code J07 for the period from 1 January 2005 to 31 December 2017. Seven-, 10- and 13- valent pneumococcal conjugate vaccines were included. At the end of the observational period, some children in the 2016 birth-cohort had not yet reached the age during which it is common to receive the third vaccine dose.

Birth-cohort	Zero doses	One dose	Two doses	Three doses
2005	4,207 (99.6)	10 (0.2)	5 (0.1)	4 (0.1)
2006	4,278 (99.1)	26 (0.6)	8 (0.2)	3 (0.1)
2007	4,345 (98.1)	51 (1.2)	18 (0.4)	13 (0.3)
2008	4,348 (94.8)	140 (3.1)	62 (1.4)	37 (0.8)
2009	4,292 (89.8)	166 (3.5)	237 (5.0)	87 (1.8)
2010	3,660 (77.8)	158 (3.4)	336 (7.1)	549 (11.7)
2011	263 (5.9)	44 (1.0)	144 (3.3)	3,976 (89.8)
2012	199 (4.5)	45 (1.0)	154 (3.5)	4,059 (91.1)
2013	165 (3.9)	44 (1.0)	122 (2.9)	3,940 (92.3)
2014	127 (2.9)	54 (1.2)	191 (4.4)	3,978 (91.4)
2015	70 (1.7)	60 (1.5)	283 (6.9)	3,672 (89.9)
2016	45 (1.1)	76 (1.9)	466 (11.4)	3,514 (85.7)

Some children in vaccine non-eligible cohorts received one, two or three doses of pneumococcal conjugate vaccines before four years of age. This generally occurred at an older age than children in the vaccine eligible cohorts Figure 4.4.

Figure 4.4: The age at which children receive their first, second and third dose of a pneumococcal conjugate vaccine (PCV) is shown as a function of birth-date. All administered doses of PCV were extracted from the National Vaccine Registry of the Icelandic Directorate of Health from 1 January 2005 to 31 December 2017, using Anatomical Therapeutic Chemical code J07. The age at which each individual child received a dose of PCV is illustrated with a point. The child’s first documented dose is indicated in red color, their second dose in blue and their third dose in green. The introduction of PCV10 into the Icelandic pediatric vaccination program on 1 April 2011, is shown with a dotted line. Though some children received PCV prior to the vaccine introduction, they did so sporadically and few achieved two or more doses. Those who did were almost invariably born in the later half of 2010. An obvious change occurs after vaccine introduction. The apparent oblique cut-off in points starting for children born in 2014 is explained by censored follow-up time.

4.1.5 The National Drug Prescription Registry

The National Drug Prescription Registry (NDPR) contains all filled outpatient prescriptions from 2005-2017. From this registry, all antibacterials for systemic use (ATC code J01 and sub-levels), vaccines (J07 and sub-levels), opthalmologicals (S01 and sub-levels) and otologicals (S02 and sub-levels) were extracted. A total of 4,020,624 prescriptions were recorded among 360,560 individuals. The number of prescriptions by the chemical sub-levels of the ATC classification system are shown in Table 4.5. The highest number of antimicrobial prescriptions filled by a single individual was 336 during the study period.

Table 4.5: The number of filled outpatient prescriptions by Anatomical Therapeutic Chemical (ATC) is shown for the period from 1 January 2005 to 31 December 2017. Data were extracted from the National Drug Prescription Registry of the Icelandic Directorate of Health.

ATC chemical sub-level code	Description	No of prescriptions
J01A	Tetracyclines	357,498
J01B	Amphenicols	0
J01C	Beta-lactam antibacterials, penicillins	1,720,661
J01D	Other beta-lactam antibacterials	106,757
J01E	Sulfonamides and trimethoprim	168,045
J01F	Macrolides, lincosamides and streptogramins	344,098
J01G	Aminoglycoside antibacterials	71
J01M	Quinolone antibacterials	135,864
J01R	Combinations of antibacterials	0
J01X	Other antibacterials	96,318
J07A	Bacterial vaccines	9,687
J07B	Viral vaccines	16,703
J07C	Bacterial and viral vaccines	496
J07X	Other vaccines	0
S01A	Anti-infective opthalmologicals	287,904
S02A	Anti-infective otologicals	1
S01C	Anti-inflammatory agents and anti-infectives opthalmologicals	40,315
S02C	Anti-inflammatory agents and anti-infectives otologicals	25,218

The distribution in the number of prescriptions for selected chemical sub-levels of the ATC classification system are shown as a function of age in Figure 4.5.

Figure 4.5: The number of filled antimicrobial prescriptions are shown as a function of age for selected chemical sub-levels of the Anatomical Therapeutic Chemical (ATC) classification system. Data were extracted from the National Drug Prescription Registry of the Icelandic Directorate of Health for the period of 1 January 2005 to 31 December 2017. Different Y-axis scales are used for each category. The figure demonstrates a similar age-distribution for beta-lactam antibacterials, sulfonamides and macrolides, with most prescriptions being filled by young children. Tetracyclines and quinolones are more commonly filled by adults.

4.1.6 Reimbursement database of Icelandic Health Insurance

All interactions with independent health care practitioners were recorded in Icelandic Health Insurance’s reimbursement database. From this database, all records of otolaryngological procedures were extracted. A total of 51,814 procedures were recorded among 34,084 individuals (Table 4.6). In total, 16,096 tonsillectomies and 29,689 tympanostomy tube placements were performed. The absolute number of adenoidectomies performed in Iceland cannot be deduced from the reimbursement database as the reimbursement codes for tonsillectomies are the same whether or not an adenoidectomy was performed concurrently.

Table 4.6: The number of procedures reimbursed to independently practicing otolaryngologists from 1 January 2005 to 31 December 2016. Data were extracted from the reimbursement database of Icelandic Health Insurance. Tympanostomy tube placements for which an anesthesiologist also received reimbursement were performed under general anesthesia. Others were categorized as TTP without mention of anesthetic. The total number of adenoidectomies performed in Iceland cannot be deduced from the reimbursement data, because the same reimbursement code is used for tonsillectomies with or without adenoidectomies. However, a seperate reimbursement code always exists for procedures including TTP.

Procedure	No of procedures
Adenoidectomy	2,442
Adenoidectomy and TTP	10,849
Myringoplasty	135
Myringotomy under local anesthetic	1,004
Tonsillectomy (+/- adenoidectomy)	9,383
Tonsillectomy (+/- adenoidectomy) and TTP	31
Tonsillectomy performed with laser (+/- adenoidectomy)	4,996
Tonsillectomy performed with laser (+/- adenoidectomy) and TTP	686
TTP under general anesthesia	16,829
TTP without mention of anesthetic	294
Tympanostomy tube removal	5,165

4.2 Impact on otitis media with treatment failure (Paper I)

The total number of children under 18 years of age who lived within Children’s Hospital Iceland’s referral region remained stable during the study period from 1 January 2008 to 31 December 2015, decreasing from 62,067 in to 61,798. The variation in the number of children under four years of age in the same region was more pronounced, increasing from 13,562 in 2008 to 14,644 in 2011, and then decreasing again to 13,272 in 2015.

During the study period, 103,220 visits were recorded to the emergency department of Children’s Hospital Iceland. The visits varied over the calendar year, spiking in the winter months and troughing in the summer months. The total number of visits increased steadily during the study period, from 12,229 in 2008 to 14,502 in 2015 (Figure 4.6).

Figure 4.6: The monthly number of visits of children zero to 18 years of age to Children’s Hospital Iceland’s emergency department during the period from January 2002 to December 2016. All visits are included regardless of International Classification of Diseases, 10th revision (ICD-10) diagnostic codes. The study period was from January 2008 to December 2015, and is delineated with two vertical dashed lines.

During the same period, 6,232 visits to the Children’s Hospital Iceland for acute otitis media were recorded for 4,624 individual children under four years of age, representing 4,994 distinct episodes. Of those episodes, 531 were treated with one or more doses of ceftriaxone. The total number of visits, visits for AOM and ceftriaxone treatment episodes are shown in Table 4.7,

Table 4.7: The incidence of visits and ceftriaxone treatment episodes at Children’s Hospital Iceland by calendar-year. The incidence is both presented for all visits regardless of diagnosis (Total) and visits associated with acute otitis media (AOM). When presented for total visits or treatment episodes regardless of diagnosis, the denominator is children 18 and younger who live within Children’s Hospital Iceland’s referral region and is expressed per 1,000 visits or person-years. The denominator of AOM associated visits or treatment episodes are children younger than four years of age in the same region. The incidence rate and incidence risk are shown with the number of events within parentheses

Year	Total (n)	AOM (n)	Total (n)	AOM (n)	Total (n)	AOM (n)
2008	197 (12,229)	69 (936)	80.8 (988)	186 (174)	15.9 (988)	72.9 (174)
2009	199 (12,514)	72 (1,012)	74.8 (936)	192 (194)	14.9 (936)	66.5 (194)
2010	181 (11,339)	64.2 (925)	81 (918)	253 (234)	14.6 (918)	63.7 (234)
2011	201 (12,645)	60.8 (890)	63.8 (807)	178 (158)	12.8 (807)	55.1 (158)
2012	215 (13,150)	58.4 (830)	52.5 (691)	163 (135)	11.3 (691)	48.6 (135)
2013	221 (13,518)	55.2 (772)	54.7 (739)	105 (81)	12.1 (739)	52.8 (81)
2014	216 (13,323)	52 (708)	48.9 (652)	76.3 (54)	10.6 (652)	47.9 (54)
2015	235 (14,502)	55.1 (731)	56.7 (822)	88.9 (65)	13.3 (822)	61.9 (65)

The incidence rate of AOM visits to Children’s Hospital Iceland decreased significantly in the post-vaccine period as compared to the pre-vaccine period; from 47.4 visits to 41.8 per 1,000 person-years. The crude IRR was 0.88 (95% CI 0.83 to 0.93). Mantel-Haenszel adjustment was not appropriate due to effect heterogeneity (\(\chi^2\) = 15.2, P<0.001). When each age-group was examined separately, a significant decrease in AOM visits was observed among children between one and two years of age (IRR 0.89) and between two and three years of age (IRR 0.79) as shown in Table 4.8. Children younger than one year of age and children between three and four years of age, visited the Children’s Hospital Iceland because of AOM 471 times and 379 times, respectively.

Table 4.8: The incidence rate ratios (IRR) of acute otitis media (AOM) associated visits between the pre- and post-vaccine periods are presented for each age with 95% confidence intervals within parentheses. The chi-squared statistic and P-value are also presented.

Age (years)	IRR (95% CI)	Chi-squared	P-value
<1	1.10 (0.90-1.30)	0.80	0.37000
1-2	0.89 (0.83-0.96)	8.60	0.00341
2-3	0.79 (0.71-0.88)	17.00	< 0.001
3-4	1.00 (0.85-1.30)	0.22	0.63900

Independent of the decrease in AOM associated visits to the Children’s Hospital, the incidence of ceftriaxone treatment episodes for AOM was also found to decrease significantly in the post-vaccine period compared to the pre-vaccine period. The effect was heterogeneous across age-strata (\(\chi^2\) = 57, P<0.001) and the crude overall IRR was 0.48 (95% CI 0.40 to 0.58). The stratum specific results are shown in Table 4.9. During the study period, only 17 episodes of AOM were treated with ceftriaxone among children zero to one years of age and 19 episodes were treated among children three to four years of age.

Table 4.9: The incidence rate ratios (IRR) of ceftriaxone treatment episodes of acute otitis media (AOM) between the pre- and post-vaccine periods are presented for each age with 95% confidence intervals within parentheses. The chi-squared statistic and P-value are also presented.

Age (years)	IRR (95% CI)	Chi-squared	P-value
<1	0.61 (0.19-1.80)	0.96	0.326
1-2	0.47 (0.37-0.60)	41.00	<0.001
2-3	0.47 (0.32-0.68)	18.00	<0.001
3-4	0.85 (0.31-2.30)	0.12	0.732

The risk of receiving ceftriaxone treatment if presenting to Children’s Hospital Iceland with AOM was calculated in order to correct for the possibility that observed decreases in ceftriaxone treatment episodes were due only to a decrease in the number of AOM associated visits. The risk decrease was not homogeneous across age-strata (\(\chi^2\) = 33.8, P<0.001) and the overall relative risk ratio was 0.58 (95% CI 0.48 to 0.69). The stratum specific effects are shown in Table 4.10.

Table 4.10: The incidence risk ratio (IRR) of receiving ceftriaxone treatment if presenting to Children’s Hospital Iceland with acute otitis media (AOM) between the pre- and post-vaccine periods is shown along with 95% confidence intervals. The corresponding Chi-squared statistic and P-value are also presented.

Age (years)	IRR (95% CI)	Chi-squared	P-value
<1	0.56 (0.17-1.70)	1.30	0.25800
1-2	0.53 (0.42-0.67)	26.00	< 0.001
2-3	0.59 (0.40-0.86)	7.50	0.00607
3-4	0.81 (0.29-2.20)	0.19	0.66200

Thus the study found significant decreases in the incidence of AOM visits, ceftriaxone treatment episodes of AOM and risk of ceftriaxone treatment if presenting to the Children’s Hospital Iceland with AOM. Similar decreases were established in the ceftriaxone treatment episodes for pneumonia. In the pre-vaccine period, 251 treatment episodes were recorded, compared to only 90 in the post-vaccine period. The effect was not consistent across age-strata (\(\chi^2\) = 72, P<0.001). The overall IRR was 0.37 (95% CI 0.29 to 0.47). The stratum specific effects are shown in Table 4.11.

Table 4.11: The incidence rate ratio (IRR) of ceftriaxone treatment episodes of pneumonia between the pre- and post-vaccine periods are presented along with 95% confidence intervals. The Chi-squared statistic and P-value are also shown.

Age (years)	IRR (95% CI)	Chi-squared	P-value
<1	0.15 (0.017-0.64)	8.6	0.00345
1-2	0.34 (0.220-0.51)	33.0	< 0.001
2-3	0.36 (0.230-0.54)	28.0	< 0.001
3-4	0.51 (0.290-0.89)	6.4	0.01170

To ascertain whether a decrease in ceftriaxone use occurred in vaccinated children for non-vaccine related indications, the incidence of ceftriaxone treatment episodes for all other indications was examined. No heterogeneity across age-strata was detected (\(\chi^2\) = 0.56, P=0.455). The Mantel-Haenszel adjusted IRR was 0.96 (95% CI 0.87 to 1.06), and the null hypothesis of no difference in the incidence rate of treatment episodes could not be rejected. The number of treatment episodes by age and vaccine period ranged from 117 to 295. The stratum specific IRR are shown in Table 4.12.

Table 4.12: The incidence rate ratios (IRR) of ceftriaxone treatment episodes for indications other than acute otitis media and pneumonia between the pre- and post-vaccine periods are presented for each age with 95% confidence intervals within parentheses. The chi-squared statistic and P-value are also presented.

Age (years)	IRR (95% CI)	Chi-squared	P-value
<1	1.30 (1.10-1.50)	7.60	0.00597
1-2	0.86 (0.70-1.00)	2.40	0.12100
2-3	0.73 (0.58-0.91)	8.00	0.00473
3-4	0.90 (0.70-1.20)	0.62	0.43200

The quarterly incidence of ceftriaxone treatment episodes by indication are shown in Figure 4.7.

Figure 4.7: The quarterly incidence rate (IR) of ceftriaxone treatment episodes are shown stratified by age and indication for the period from January 2008 to December 2015. The IR per 1,000 person-years is presented for children one, two, three and four years of age using different line-types. The figure is stratified by indication, and demonstrates a decrease in the IR of ceftriaxone treatment episodes for acute otitis media and pneumonia in the post-vaccine period January 2012 to December 2016. No such decrease is visible for ceftriaxone treatment episodes for all other indications.

To further test whether a general decrease was occurring in the overall use of ceftriaxone, rather than a specific decrease for vaccine-related indications in vaccinated children, an examination of ceftriaxone treatment episodes in all children regardless of age and indication was undertaken. An overall decrease in the IR of ceftriaxone treatment episodes was found among children under 18 years of age regardless of indication. The IR declined from 11.1 to 9.55 treatment episodes per 1,000 person-years, IRR 0.86 (95% CI 0.81-0.91). The effect was not consistent across age-groups (\(\chi^2\) = 23.6, P<0.001). When examined by age-group, the overall decrease proved to be driven by the youngest age-group – i.e the children who were protected by the vaccination. The incidence of ceftriaxone treatment episodes did not decrease significantly in other age groups (Figure 4.8).

Figure 4.8: The monthly incidence rates (IR) of ceftriaxone treatment episodes per 1,000 person-years regardless of indication are shown stratified by age-groups for the period from January 2005 to December 2016. The delineation between the pre- and post-vaccine periods on January 2012 is illustrated with a dashed vertical line. For each age-group, the mean IR in the pre- and post-vaccine periods are depicted with a solid horizontal line. The incidence rate ratio (IR) between the two vaccine periods is shown for each age-group along with 95% confidence intervals within parentheses. The figure demonstrates a significant decrease in the IR of ceftriaxone treatment episodes for children zero to three years of age. No change is detectable in other age-groups.

4.3 Impact on primary care visits for acute otitis media (Paper II)

The demographics of the study birth-cohorts are described in chapter 4.1 and Table 4.1. A total of 92,935 primary care visits for acute otitis media were recorded among birth-cohorts 2005-2015 during the study period from 1 January 2005 to 31 December 2015. The crude incidence rate of AOM visits to primary care per 100 person-years in the VNEC and VEC was 45.3 and 39.8 respectively. The IR and number of AOM visits by birth-cohort and gender are shown in Table 4.13.

Table 4.13: The incidence rate (IR) of acute otitis media visits to primary care physicians is shown for each birth-cohort and gender. The absolute number of visits are presented within parentheses.

Birth-cohort	Females	Males
2005	41.9 (2,777)	49.0 (3,439)
2006	46.1 (3,096)	50.9 (3,605)
2007	45.7 (3,118)	50.3 (3,646)
2008	46.2 (3,259)	45.3 (3,419)
2009	40.9 (2,981)	47.0 (3,649)
2010	45.0 (3,207)	47.0 (3,523)
2011	39.1 (2,631)	44.1 (3,164)
2012	40.6 (2,760)	41.8 (2,977)
2013	38.0 (2,125)	42.8 (2,322)
2014	37.4 (1,200)	44.0 (1,465)
2015	15.8 (157)	20.8 (222)

The lowest incidence was observed in children zero to three months of age, ranging from 3-6 visits per 100 person-years. Thereafter, the incidence increased sharply, and peaked in children eight to eleven and twelve to fifteen months of age, ranging from 50 to 80 visits per 100 person-years. The crude IR decreased significantly in all age-groups, with incidence rate ratios ranging from 0.60 to 0.94. The largest and visually most consistent decrease in incidence was observed among children zero to three months of age, IRR 0.6 (95%CI 0.51 to 0.69), Figure 4.9.

Figure 4.9: The incidence rate (IR) of acute otitis media (AOM) visits to primary care physicians is shown stratified by birth-cohort and four-month age-groups. The estimated IR is illustrated with a point and 95% confidence intervals are depicted with horizontal error-bars. The vaccine non-eligible birth-cohorts (VNEC) are illustrated in red, and the vaccine eligible birth-cohorts in blue. The incidence rate ratio (IRR) between the VEC and VNEC is shown for each age-group and 95% confidence intervals are presented within parentheses.

When tabulated by the cumulative number of AOM episodes experienced by each child, the proportion of children experiencing zero episodes of AOM increased in the VEC compared to the VNEC, while the proportion experiencing one to four episodes and five or more decreased, as shown in Table 4.14.

Table 4.14: The proportion of each vaccine eligibility cohort that experienced zero, one to four and five to twelve cumulative episodes of AOM by 36 months of age. The incidence risk ratio (IRR) between the vaccine eligible cohorts (VEC) and vaccine non-eligible cohorts (VNEC) is presented along with 95% confidence intervals within parentheses.

No. visits	VNEC (%)	VEC (%)	Incidence risk (95%CI)
0	40.0	43.2	1.14 (1.10-1.18)
1-4	55.7	53.2	0.904 (0.876-0.932)
5-12	4.23	3.58	0.84 (0.744-0.946)

Discrimination indices for the Andersen-Gill multiple event model were adequate, Nagelkerke’s \(R^2\) = 0.110 and Somer’s \(D_{xy}\) = 0.238. No systematic deviations in Schoenfeld residuals were detected on diagnostic plots indicating that the proportional hazard assumption for each covariate were met. There was little variation in the hazard of AOM between vaccine non-eligible birth-cohorts. Only the 2007 birth-cohort differed significantly, with a hazard ratio of 1.06 (95%CI 1.01 to 1.12) compared to the 2010 birth-cohort. An abrupt and significant decrease in the hazard of AOM was noted in the first vaccine eligible cohort, which continued for all remaining VEC (Figure 4.10). The estimated impact of PHiD-CV10 on AOM episodes in the primary care setting among children younger than three years of age was 21% (95%CI 11% to 30%).

Figure 4.10: The hazard ratio (HR) of acute otitis media (AOM) between each birth-cohort and the last vaccine non-eligible birth-cohort is shown. The estimated HR is presented with a point, and 95% confidence intervals are illustrated with horizontal error-bars. The 2010 birth-cohort is used as a reference and therefore no confidence intervals are presented. A dashed vertical line is placed on the ratio value of one to assist in visually estimating significance. The X-axis is on the logarithmic scale. The figure demonstrates an abrupt decrease in the hazard of AOM in the first vaccine eligible birth-cohort.

When the hazard ratio of AOM between VEC and VNEC was stratified by the number of previous AOM episodes, the vaccine impact was discernible in children who had experienced either no or only one previous AOM episode. Among children who had more than one previous AOM episode, no effect was found (Figure 4.11).

Figure 4.11: The hazard ratio (HR) of acute otitis media (AOM) between the vaccine eligible (VEC) and vaccine non-eligible cohorts (VNEC) is shown stratified by the number of previous AOM episodes. The estimated HR is illustrated as a solid black line, and 95% confidence intervals are presented as a shaded area. A dashed horizontal line is placed on the ratio value of one to assist in visually assessing significance. The Y-axis is truncated at a HR of 0.5 and is presented on the logarithmic scale. The figure demonstrates a significantly lower hazard of experiencing an additional episode of AOM among vaccine eligible children who have previously experienced zero or one episodes.

The mean number of AOM episodes in primary care was calculated as a function of age using the generalized Nelson-Aalen estimate of the underlying Andersen-Gill model. By their third birthday, the average child in the VNEC had experienced 1.61 episodes of AOM. The average child in the VEC had experienced 1.37. The mean number of AOM episodes by age is shown in Figure 4.12.

Figure 4.12: The mean number of acute otitis media (AOM) episodes is shown as a function of age. The estimated mean is presented as a solid red line for the vaccine non-eligible cohorts and a blue line is used for the vaccine eligible cohorts. The 95% confidence intervals are illustrated as shaded areas. The figure demonstrates an early divergence in the mean number of episodes between the two cohorts.

4.4 Impact on outpatient antimicrobial prescriptions (Paper III)

Demographic data regarding the study birth-cohorts are summarized in chapter 4.1 and Table 4.1. During the study period from 1 January 2005 to 31 December 2016, a total of 276,109 prescriptions were filled for 55,599 Icelandic children under three years of age. From 2005-2012, first-line penicillins were the most commonly prescribed antimicrobials and represented between 41% and 47% of all antimicrobial prescriptions in this age-group. Their use decreased suddenly to 32% in 2013, and represented only 18% of all antimicrobial prescriptions in 2014 and 2015. During this same period, the use of second-line penicillins increased from 35%-40% from 2005-2012, to 48%, 55% and 54% in 2013, 2014 and 2015. Use of cephalosporins followed a similar trend – their use represented between 5.2% and 7.8% of all prescriptions 2005–2012, and increased to 10–15% between 2013–2016. Antimicrobial prescriptions by calendar year are shown in Table 4.15.

Table 4.15: The incidence rate (IR) of outpatient antimicrobial prescriptions among children younger than three years of age is shown by calendar-year. Antimicrobials were grouped based on a previously published classification system (Youngster et al. 2017). First-line penicillins include amoxicillin, dicloxacillin and phenoxymethylpenicillin. Amoxicillin-clavulanate is the only antimicrobial included in the second-line penicillin category and erythromycin is the only first-generation macrolide. Second-generation macrolides include azythromycin and clarithromycin. Cefuroxime and cephalexine are the most common cephalosporins and the category Other includes mostly trimethoprim and trimethoprim and sulfamethoxazole.

Calendar year	IR (n)	1st-line penicillin	2nd-line penicillin	1st-gen macrolide	2nd-gen macrolide	Cephalosporin	Other
2005	204 (12,570)	41.4	37.9	1.5	6.5	5.4	7.3
2006	206 (12,844)	40.3	39.6	1.3	6.2	5.4	7.2
2007	192 (13,111)	45.0	36.8	1.6	6.4	5.2	5.1
2008	178 (13,474)	46.7	35.2	0.2	6.4	5.9	5.6
2009	159 (14,062)	46.4	37.2	0.0	5.5	6.3	4.5
2010	167 (14,382)	43.7	38.5	0.0	5.5	7.0	5.2
2011	164 (14,588)	44.7	37.9	0.0	5.9	7.5	4.0
2012	160 (14,225)	43.5	39.0	0.0	6.9	7.8	2.8
2013	152 (13,893)	32.1	48.1	0.0	6.6	10.0	3.2
2014	152 (13,390)	18.5	55.5	0.0	6.6	14.5	4.9
2015	150 (13,284)	18.5	53.9	0.1	7.3	15.0	5.3
2016	160 (12,813)	35.3	41.7	0.0	5.5	12.9	4.6

The proportion of visits resulting in antimicrobial prescription and the incidence of antimicrobial prescriptions linked to each of the study’s diagnostic groups are shown in Figure 4.13. The proportion of otitis media visits resulting in an antimicrobial prescription remained stable at between 57% and 64% of visits. The incidence of otitis media associated prescriptions decreased from a high of 54.9 prescriptions per 100 person-years in 2008 to 39.8 prescriptions per 100 person-years in 2015.

Figure 4.13: The proportion of visits resulting in antimicrobial prescription and the incidence of prescriptions are depicted for each diagnostic group for the period from 2005 to 2015. The diagnostic groups and their corresponding color are presented in the figure legend. The figure demonstrates that the proportion of visits resulting in antimicrobial prescriptions remains fairly stable for most diagnostic groups, while the incidence is decreasing.

During the study period, a total of 226,084 outpatient antimicrobial prescriptions were recorded among birth-cohorts 2005-2015. The crude incidence rate of outpatient antimicrobial prescriptions per 100 person-years in the VNEC and VEC was 164.6 and 150.2 respectively. The incidence rate and number of outpatient antimicrobial prescriptions by birth-cohort and gender is shown in Table 4.16.

Table 4.16: The incidence rate (IR) of outpatient antimicrobial prescriptions is presented by birth-cohort and gender. The absolute number of prescriptions are shown within parentheses.

Birth-cohort	Females	Males
2005	176.0 (11,178)	200.0 (13,423)
2006	167.0 (10,843)	190.0 (13,109)
2007	153.0 (10,140)	174.0 (12,339)
2008	153.0 (10,543)	171.0 (12,492)
2009	151.0 (10,699)	169.0 (12,775)
2010	150.0 (10,366)	161.0 (11,854)
2011	142.0 ( 9,230)	156.0 (10,906)
2012	142.0 ( 9,447)	158.0 (11,058)
2013	138.0 ( 9,015)	158.0 (10,180)
2014	145.0 (7,726)	167.0 (9,234)
2015	138.0 (4,075)	173.0 (5,452)

When stratified by six month age-groups, the lowest incidence was observed in children zero to five months of age and ranged from 30 to 50 prescriptions per 100 person-years. The observed incidence increased sharply thereafter and peaked among children twelve to seventeen months of age, ranging from 225 to 280 prescriptions per 100 person-years. The crude IR decreased significantly in all age-groups, with incidence rate ratios ranging from 0.82 to 0.94 (Figure 4.14).

Figure 4.14: The incidence rate (IR) of outpatient antimicrobial prescriptions is shown stratified by birth-cohort and six-month age-groups. The estimated IR is shown as a point and 95% confidence intervals are illustrated with horizontal error-bars. The vaccine non-eligible cohorts (VNEC) are depicted in red, and the vaccine eligible cohorts (VEC) in blue. The incidence rate ratio (IRR) between the VEC and VNEC is shown for each age-group and 95% confidence intervals are presented within parentheses.

The proportion of children in the VNEC and VEC who filled at least one antimicrobial prescription by three years of age was 88.6% and 86.8 respectively. Children in the VEC were significantly more likely than children in the VNEC not to have filled an antimicrobial prescription (incidence risk ratio 1.16, 95%CI 1.10 to 1.23) or to have filled only between one and four antimicrobial prescriptions (incidence risk ratio 1.08, 95%CI 1.06 to 1.11). The cumulative number of prescriptions by vaccine eligibility cohort in shown in Table 4.17.

Table 4.17: The proportion of children in the vaccine non-eligible cohorts (VNEC, 2005–2010) and vaccine eligible cohorts (VEC, 2011–2013) who filled 0, 1–4, 5–9, 10–14 and ≥15 antimicrobial prescriptions by 36 months of age.

No. prescriptions	VNEC (%)	VEC (%)	Incidence risk (95%CI)
0	11.4	13.2	1.16 (1.10-1.23)
1-4	43.7	47.3	1.08 (1.06-1.11)
5-9	31.6	29.0	0.918 (0.889-0.947)
10-14	9.79	7.52	0.768 (0.716-0.823)
≥15	3.51	2.91	0.831 (0.74-0.934)

Discrimination indices for the Andersen-Gill multiple event model were adequate, Nagelkerke’s \(R^2\) = 0.212 and Somer’s \(D_{xy}\) = 0.295, and no significant deviations from the model assumptions were visible on diagnostic plots. The model was used to estimate the hazard ratio of outpatient antimicrobial prescriptions between each of the study’s birth-cohorts and the last vaccine non-eligible cohort, 2010. Visually, there seemed to be a decreasing trend in hazard of prescription among the vaccine non-eligible birth-cohorts (Figure 4.15). The hazard did not change significantly between the last vaccine non-eligible birth-cohort and the preceding two cohorts. It did decrease significantly thereafter, with each vaccine eligible cohort having a significantly lower hazard of outpatient antimicrobial prescription. The estimated impact of PHiD-CV10 on outpatient antimicrobial prescriptions among children younger than three years of age was 8% (95%CI 4% to 12%).

Figure 4.15: The hazard ratio (HR) of outpatient antimicrobial prescription between each birth-cohort and the last vaccine non-eligible birth-cohort is shown. The estimated HR is depicted as a point, and 95% confidence intervals are illustrated with horizontal error-bars. The 2010 birth-cohort is used as a reference and by definition, no uncertainty is present. Therefore, no confidence intervals are presented. A dashed vertical line is placed on the ratio value of one to assist in visually estimating significance. The X-axis is on a logarithmic scale. The figure demonstrates a decreasing trend in the hazard of outpatient antimicrobial prescription in the vaccine non-eligible cohorts. An abrupt decrease in the hazard is observed in the first vaccine eligible birth-cohort.

When stratified by the number of previous prescriptions, an independent vaccine impact on subsequent prescriptions was still discernible in children who had received up to three prior antimicrobial prescriptions. Among children who had received more than three prior prescriptions, no effect was found (Figure 4.16).

Figure 4.16: The hazard ratio (HR) of outpatient antimicrobial prescriptions between the vaccine eligible (VEC) and vaccine non-eligible cohorts (VNEC) is shown stratified by the number of previous prescriptions. The estimated HR is depicted as a solid black line, and 95% confidence intervals are illustrated as a shaded area. A dashed horizontal line is placed on the ratio value of one to assist in visually assessing significance. The Y-axis is truncated at a HR of 0.5 and is presented on a logarithmic scale. The figure demonstrates a significantly lower hazard of filling additional outpatient antimicrobial prescriptions among vaccine eligible children who had previously filled fewer than four prescriptions.

The mean number of outpatient antimicrobial prescriptions as a function of age was calculated using the generalized Nelson-Aalen estimate of the underlying Andersen-Gill model. The average male child in the VNEC had filled 6.48 antimicrobial prescriptions by his fourth birthday, and the average female child had filled 6.07. The average male and female children in the VEC had filled 5.84 and 5.46 prescriptions respectively. The mean number of antimicrobial prescriptions by age and gender is shown in Figure 4.17.

Figure 4.17: The mean number of outpatient antimicrobial prescriptions are shown as a function of age, and stratified by vaccine eligibility cohort and gender. The figure deviates from the color schema used in other figures that compromise this thesis. Here, the estimates for females are illustrated in red and males in blue. Vaccine non-eligible cohorts are represented with solid lines and vaccine eligible cohorts with dashed lines. The 95% confidence intervals are illustrated as shaded areas. The figure demonstrates an early divergence in the mean number of prescriptions, and a consistent difference between genders.

4.5 Impact on tympanostomy tube procedures (Paper IV)

Demographic data regarding the study birth-cohorts are summarized in chapter 4.1. In total, during the study period from 1 January 2005 to 31 December 2016, 14,351 children underwent 20,373 tympanostomy tube placements, 57% of whom were male.

The median age of children undergoing their first tympanostomy procedure was 17 months (IQR 13-24, 18% younger than one year of age). In the subset of children who underwent a TTP during the study period, 10,248 (71%) underwent only one procedure, 2,902 (20%) underwent two, and 1201 (8%) underwent three or more. Almost all (98%) of the procedures were performed in private outpatient clinics. The number of otolaryngologists performing outpatient TTP increased from 15 in 2005 to 23 in 2016. Each surgeon performed a median of 123 (IQR: 56.5-196) procedures each year. The study’s population is summarized in Table 4.18.

Table 4.18: An overview of the birth-cohorts included in paper IV. The number of observed tympanostomy tube placements (TTP) in each cohort is presented, with the number of children undergoing the procedures within parentheses. The median age in months at the time of a childs first procedure is shown along with the interquartile range. The observational period is 1 January 2005 to 31 December 2016. Birth-cohorts 2012-2015 do not attain full follow-up time, as indicated by fewer person-years included in the study. Thus, the decrease in the number of procedures and and median age is not indicative of a true decrease.

Birth-cohort	Number of children	Person-years	Number of procedures (n children)	Median age (months)
2005	4,541	21,409	1,946 (1,280)	17 (12-25)
2006	4,665	21,988	1,931 (1,303)	18 (13-27)
2007	4,770	22,500	1,974 (1,335)	18 (13-27)
2008	4,949	23,313	2,140 (1,428)	18 (13-26)
2009	5,128	24,141	2,145 (1,514)	18 (13-25)
2010	4,984	23,580	2,203 (1,547)	18 (13-26)
2011	4,642	22,056	1,997 (1,382)	18 (13-24)
2012	4,668	20,195	2,057 (1,419)	16 (12-23)
2013	4,442	14,964	1,642 (1,200)	16 (13-23)
2014	4,444	10,744	1,582 (1,251)	16 (13-20)
2015	4,136	5,983	756 (692)	13 (11-15)

The crude incidence rate of TTP in the vaccine eligible cohorts was 10.6 procedures per 100 person-years. This was significantly higher than the crude incidence rate in the vaccine non-eligible cohorts, 8.7 procedures per 100 person-years (IRR 1.20, 95%CI 1.17 to 1.24). When stratified by age-groups, the crude incidence rate was highest among 12-17 month old children, ranging from 19 to 27 procedures per 100 person-years (Figure 4.18).

Figure 4.18: The incidence rate (IR) of tympanostomy tube placements (TTP) is shown stratified by birth-cohort and six-month age-groups. The estimated IR is represented with a point, and 95% confidence intervals are illustrated with horizontal error-bars. The vaccine non-eligible cohorts (VNEC) are depicted in red, and the vaccine eligible cohorts (VEC) in blue. The incidence rate ratio (IRR) between the VEC and VNEC is shown for each age-group and 95% confidence intervals are presented within parentheses.

The cumulative incidence of children who had undergone at least one TTP by five years of age was highest in birth-cohort 2010 (31.7%), and lowest in birth-cohort 2006 (28.6%), Table 4.19. The cumulative incidence of tympanostomy procedures was significantly higher in the VEC compared to VNEC regardless of age (Figure 4.19).

Table 4.19: The cumulative incidence of having undergone at least one tympanostomy tube placement (TTP) at six-month age intervals is presented for each birth-cohort. The cumulative incidence is presented as a percentage (%) of all children in the respective cohort. Birth-cohorts 2012-2015 do not attain full follow-up time. Lack of information due to censoring is indicated with a hyphen (-).

Age (months)	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015
6	0.4	0.3	0.4	0.2	0.3	0.1	0.4	0.2	0.3	0.2	0.5
12	7.2	7.1	6.6	7.2	6.5	6.6	6.6	7.8	6.5	6.9	7.6
18	16.4	14.8	14.9	15.9	15.7	16.5	16.3	18.3	16.3	19.1	18.4
24	21.1	19.5	19.7	21.0	22.0	22.4	23.3	23.9	21.7	26.1	-
30	23.7	22.9	23.4	24.5	25.1	26.3	26.3	26.7	24.9	29.0	-
36	25.8	24.8	25.3	26.5	27.2	28.8	27.9	28.3	26.9	-	-
42	26.9	26.5	26.7	27.7	28.5	30.1	28.7	29.5	27.5	-	-
48	27.8	27.4	27.8	28.5	29.1	31.0	29.4	30.4	-	-	-
54	28.4	28.3	28.4	29.1	29.9	31.4	30.1	30.9	-	-	-
59	28.8	28.6	28.6	29.5	30.2	31.7	30.4	31.3	-	-	-

Figure 4.19: The cumulative incidence of tympanostomy tube placements (TTP) is shown for the vaccine eligible (VEC) and vaccine non-eligible cohorts (VNEC) as a continuous function of age. Cumulative incidence is presented as a proportion of all children in the respective vaccine eligibility cohort. The estimated cumulative incidence is illustrated with a solid red line for the VNEC, and a solid blue line for the VEC, and 95% confidence intervals are depicted with a shaded area.

In the subset of children who underwent TTP, the mean (median) number of otitis media associated visits to primary care or to the pediatric emergency department was 2.05 (2) visits in the vaccine non-eligible cohorts, compared to 1.72 (1) visits in the vaccine eligible cohorts. The distribution in the number of previous visits was significantly different between the VNEC and VEC (Chi-Squared test statistic 63.8, P<.001). The proportion of children who did not have a single recorded visit prior to undergoing the procedure increased from 20.6% in the VNEC to 28.9% in the VEC, RR 1.40 (95%CI 1.28-1.54). Children in the vaccine eligible cohorts had received significantly fewer antimicrobial prescriptions prior to undergoing the procedure (Chi-Squared test statistic 53.6, P<.001). The mean (median) number of previous antimicrobial prescriptions was 3.19 (4) in the vaccine eligible cohorts compared to 3.62 (4) in the vaccine non-eligible cohorts. Children in the VEC were more likely to have never been prescribed antimicrobials prior to undergoing tympanostomy placement, RR 1.52, 95%CI 1.18-1.96). The comparison between VEC and VNEC is summarized in Table 4.20.

Table 4.20: The cumulative number of previous antimicrobial prescriptions and otitis media visits are shown for those children in the vaccine non-eligible (VNEC) and vaccine eligible cohorts (VEC) who underwent at least one tympanostomy tube placement. The proportion of each cohort who had the corresponding number of prior visits or prescriptions is shown with the absolute number of children within parentheses. The relative risk (RR) and absolute risk difference (ARD) between the VEC and VNEC is shown with 95% confidence intervals.

Cumulative number	VNEC % (n)	VEC % (n)	RR (95%CI)	ARD (95%CI	VNEC % (n)	VEC % (n)	RR (95%CI)	ARD (95%CI
0	3.43 (286)	5.22 (72)	1.18 (1.52 to 1.96)	1.79 (0.51 to 3.07)	20.60 (1,720)	28.90 (398)	1.28 (1.40 to 1.54)	8.29 (5.70 to 10.900)
1	11.60 (966)	12.80 (177)	0.95 (1.11 to 1.29)	1.26 (-0.68 to 3.19)	24.90 (2,080)	24.40 (337)	0.89 (0.98 to 1.09)	-0.45 (-2.94 to 2.040)
2	19.30 (1,610)	22.60 (311)	1.05 (1.17 to 1.30)	3.28 (0.87 to 5.68)	20.40 (1,700)	19.60 (270)	0.85 (0.96 to 1.07)	-0.86 (-3.17 to 1.450)
3-4	37.80 (3,150)	37.40 (516)	0.92 (0.99 to 1.07)	-0.39 (-3.19 to 2.41)	24.90 (2,080)	20.20 (279)	0.73 (0.81 to 0.91)	-4.64 (-7.00 to -2.290)
5-7	22.30 (1,860)	19.30 (266)	0.77 (0.86 to 0.97)	-3.01 (-5.32 to -0.70)	7.98 (666)	6.45 (89)	0.65 (0.81 to 1.00)	-1.53 (-2.99 to -0.066)
8+	5.61 (468)	2.68 (37)	0.34 (0.48 to 0.67)	-2.93 (-3.95 to -1.90)	1.25 (104)	0.43 (6)	0.15 (0.35 to 0.79)	-0.81 (-1.28 to -0.348)

A diagnostic plot of Schoenfeld residuals did not reveal deviations from the proportional hazards assumption. The hazard of undergoing TTP was considerably higher in children who had previously visited a physician for otitis media or received an antimicrobial prescription. Children who had one prior documented visit were considerably more likely to undergo the procedure than children who had no documented visits, HR of 3.12 (95%CI 2.93 to 3.32). Likewise, children who had previously filled one antimicrobial prescription were more likely to receive a tympanostomy tube than children who had received no prescription, 6.98 (95%CI 6.13 to 7.95). The hazard of tympanostomy tube placement increased gradually from birth-cohort 2005 to 2015 (Figure 4.20).

Figure 4.20: The hazard ratio (HR) of tympanostomy tube placements (TTP) is depicted between each birth-cohort and the last vaccine non-eligible birth-cohort are depicted, as estimated by three Cox regression models. In panel A, the unadjusted HR estimates are shown. Panel B shows HR estimates adjusted for antimicrobial prescriptions, and panel C illustrates the HR estimates adjusted for otitis media associated visits. Primary care visits are obtained from the Primary Care Registry of the Directorate of Health, and the observation period is restricted to December 2015. Because of this the data presented in panel C is restricted to birth-cohorts 2005-2014. The figure demonstrates an increasing hazard of TTP with each successive birth-cohort. The effect is more pronounced when adjustment is made for antimicrobial prescriptions and visits.

4.6 Impact on respiratory associated hospitalizations (Paper V)

Demographic data regarding the study birth-cohorts are summarized in chapter 4.1. In total, 51,264 children were followed for a median of 1,096 days (range 6-1,096) resulting in 142,315 person-years of follow-up time. Of those, 1,414 children were admitted to hospital 1,703 times with diagnoses compatible with the study’s diagnostic groups. The total number of hospital admissions regardless of indication was 4,842. An overview of the demographic data is presented in Table 4.21.

Table 4.21: An overview of the birth-cohorts included in paper V. The total number of hospitalizations, hospital admissions due to the study’s diagnoses, and admissions to the intensive care unit (ICU) are presented. The number of children is presented within parentheses. The proportion of all hospital admissions that were due to the study’s diagnoses is shown. The observational period is 1 January 2005 to 31 December 2017. Birth-cohorts 2014 and 2015 do not attain full follow-up time, as indicated by fewer person-years included in the study.

Birth-cohort	Number of children	Person-years	All cause admissions, n	Study admissions, n (children, n)	Proportion due to study diagnosis, %	ICU admissions, n (children, n)
2005	4,541	13,277	446	219 (160)	49.1	7 (7)
2006	4,668	13,658	415	176 (140)	42.4	10 (8)
2007	4,770	13,985	423	186 (160)	44.0	6 (5)
2008	4,953	14,472	442	117 (101)	26.5	5 (4)
2009	5,130	14,965	484	124 (109)	25.6	7 (6)
2010	4,988	14,592	384	158 (138)	41.1	7 (7)
2011	4,644	13,640	392	129 (112)	32.9	4 (4)
2012	4,668	13,753	576	196 (155)	34.0	0 (0)
2013	4,442	13,044	472	149 (119)	31.6	9 (8)
2014	4,446	10,930	431	144 (122)	33.4	6 (5)
2015	4,136	6,140	377	105 (98)	27.9	3 (3)

Of the children in the study birth-cohorts, 550 were hospitalized 660 times with ICD-10 discharge diagnoses consistent with pneumonia. In the same cohorts, 508 children were admitted 550 times with diagnoses consistent with other acute lower respiratory tract infections. In the vaccine non-eligible cohorts, the crude incidence rate of pneumonia requiring hospital admission was 4.94 per 1,000 person-years, which decreased to 4.18 per 1,000 in the vaccine eligible cohorts. The analogous crude incidence rate of hospitalizations for other acute lower respiratory tract infections was 2.94 and 5.23 per 1,000 person-years. Though the absolute number of admissions was similar between these two groups, the distribution of cases was different. The crude incidence rate of hospital admissions for pneumonia was highest in children 12-17 months of age, while the incidence rate of hospitalizations for other lower respiratory tract infections was highest in children <6 months of age (Figure 4.21). Children admitted for other acute lower respiratory tract infections were significantly younger than children admitted for pneumonia (mean age 8.0 months and 13.6 months respectively, P<.001). Using crude age-group stratified incidence rate ratios between the vaccine eligible and non-eligible cohorts, the incidence rate of pneumonia hospitalizations was found to have decreased significantly only among children 12-17 months of age, crude incidence rate ratio 0.52 (95%CI: 0.35-0.77). Using the same method, the incidence rate of hospital admissions for other acute lower respiratory tract infections admissions was found to have increased significantly among children 0-5 months of age, crude incidence rate ratio 1.50 (95%CI 1.23-1.84).

Figure 4.21: Crude incidence rates (IR) of the six diagnostic groups (A-F) per 1,000 person-years for each of the birth-cohorts. Panels A-D and F depict the IR of admissions based on the International Classification of Diseases, 10th revision (ICD-10) discharge diagnoses, while panel E depicts the IR of admissions with culture confirmed invasive pneumococcal disease (IPD), regardless of ICD-10 diagnosis. Birth-cohorts are compared in six-month age-groups which are illustrated on the Y-axis. The vaccine non-eligible cohorts (VNEC) are illustrated in red and the vaccine eligible cohorts (VEC) in blue.

A significant difference was detected in the cumulative rate of hospital admissions for both pneumonia and acute lower respiratory tract infections between the vaccine eligible and non-eligible cohorts (Figure 4.22). The hazard ratio of hospital admission for pneumonia was 0.80 (95%CI:0.67-0.95), with an E-value of 1.81 and a lower bound of 1.29. When the risk-set was restricted to children younger than 90 days and 90 days and older, respectively, the hazard ratio was 1.22 (95%CI 0.81-1.85) and 0.73 (95%CI 0.60-0.89) respectively. The hazard ratio for hospital admission due to acute lower respiratory tract infection was 1.32 (95%CI:1.14-1.53), with an E-value of 1.97 and a lower bound of 1.54. The hazard ratio was augmented when children younger than 90 days were analysed separately, HR 1.54 (95%CI 1.23-1.94). It was not significant in children 90 days and older, HR 1.18 (95%CI 0.97-1.44).

A total of 131 hospitalizations for acute upper respiratory tract infections were recorded for 123 children. During the same period, 256 children were admitted to hospital 280 times for otitis media and complications. The crude incidence rate of hospital admissions for otitis media was higher than the incidence rate of admissions for acute upper respiratory tract infections; 2.32 and 1.45 per 1,000 person-years in the vaccine eligible and vaccine non-eligible cohorts respectively, compared to 0.78 and 1.13 per 1,000 person-years. The mean age of children admitted for acute upper respiratory tract infections was 13.5 months compared to 12.8 months for children admitted for otitis media and complications. The crude incidence rate by age-group is shown in Figure 4.21. The cumulative incidence rate of hospitalization per 1000 person-years for acute upper respiratory tract infections and otitis media and complications are shown in Figure 4.22. The hazard ratio of otitis media hospitalizations between the vaccine eligible and non-eligible cohorts was 0.57 (95%CI:0.43-0.73) with an E-value of 2.9 .and a lower bound of 2.08. When restricted to children younger than 90 days of age, the hazard ratio was 0.72 (95%CI 0.33-1.57), and when evaluating children 90 days and older it was 0.55 (95%CI 0.42-0.72). The hazard ratio for hospital admission for acute upper respiratory tract infections was 1.56 (95%CI:1.11-2.19), with an E-value of 2.49 and a lower bound of 1.46. Among children younger than 90 days, and 90 days and older respectively, the hazard ratio was 3.4 (95%CI 1.72-6.90) and 1.13 (95%CI 0.75-1.71).

Figure 4.22: Kaplan-Meier cumulative event curves per 1,000 person-years for each of the diagnostic groups are displayed in panels A-F. The vaccine non-eligible cohorts (VNEC) are illustrated in red and the vaccine eligible cohorts (VEC) in blue. The 95% confidence intervals are represented with a shaded area. The Y-axis is scaled independently for each pair of diagnostic groups (A-B, C-D and E-F).

A total of 15 children were admitted to hospital 19 times for meningitis, and 61 children were admitted 63 times for sepsis. The crude incidence rate of meningitis hospitalization was 16.5 and 8.7 per 100,000 person-years in the vaccine non-eligible and vaccine eligible cohorts respectively, and the analogous crude incidence rate for sepsis hospitalizations was 38.8 and 52.3. Culture confirmed invasive pneumococcal disease was diagnosed in 37 children under three years of age in the study birth-cohorts. Of those, 23 (59%) were admitted for inpatient treatment. Of the admitted children, eight children had a primary discharge diagnosis of Sepsis due to Streptococcus pneumoniae (A40.3), eight were diagnosed with Pneumococcal meningitis (G00.1), two with Pneumonia due to Streptococcus pneumoniae (J13), two with Bacterial pneumonia, not elsewhere classified (J15) and the remaining three were diagnosed with Bacterial meningitis, unspecified (G00.9), Pyogenic arthritis, unspecified (M00.9) and Fever, unspecified (R50.9). The crude incidence of invasive pneumococcal disease, regardless of whether the child was admitted to hospital, was 24.7 per 100,000 person-years in the VNEC compared to 1.74 per 100,000 person-years in the VEC. When only considering hospitalized invasive pneumococcal disease, the crude IR was 24.7 and 1.74 per 100,000 person-years. No vaccine-type invasive pneumococcal disease was diagnosed in the VEC. Crude incidence rates of hospitalization by age-group are shown in Figure 4.21.

Table 4.22: Hazard ratios (HR) between the vaccine eligible (VEC) and vaccine non-eligible birth-cohorts (VNEC) for each disease-group. A HR lower than one indicates a relative decrease in disease-group in the VEC compared to the VNEC, while a HR higher than one indicates an increase. A HR was not calculated for vaccine-type invasive pneumococcal disease as not cases were diagnosed and the VEC.

Disease group	Hazard ratio (95%CI)
Otitis Media and Complications	0.56 (0.44-0.73)
Acute upper respiratory infection	1.55 (1.10-2.18)
Pneumonia	0.80 (0.67-0.95)
Acute Lower Respiratory Tract Infections	1.32 (1.14-1.53)
Sepsis	1.26 (0.74-2.12)
Invasive Pneumococcal Disease	0.07 (0.01-0.50)

The mean age of children admitted for meningitis, sepsis and invasive pneumococcal disease was 9.7 months, 8.4 months and 14.4 months respectively. The cumulative incidence rates of hospitalization per 1000 person-years for sepsis and invasive pneumococcal disease are depicted in Figure 4.22. The hazard ratio of hospitalization for meningitis between the vaccine eligible and non-eligible cohorts was 0.45 (95%CI 0.15-1.41). An E-value was not computed as the hazard ratio was not significant. The hazard ratio for hospital admissions due to invasive pneumococcal disease between the vaccine eligible and vaccine non-eligible cohorts was 0.07 (95%CI:0.01-0.50), with an E-value of 28.06. and a lower bound of 3.41. The hazard ratio of a sepsis hospitalization between the vaccine eligibility cohorts was 1.26 (95%CI:0.75-2.13). No E-value was calculated as the ratio was not significant. Restricted analyses in these three diagnostic groups did not alter results significantly.

4.7 Impact and cost-effectiveness analysis (Paper VI)

4.7.1 Population impact on acute otitis media among children younger than 20 years of age

From 1 January 2005 to 31 December 2015, children younger than 20 years of age visited primary care physicians 164,453 times for acute otitis media and its complications. Strong seasonal variation was detected, with more visits occurring in December through March, and few visits occurring in June and July (Panels A and B of Figure 4.23). The monthly number of AOM visits during the post-vaccine period was lower than average in all age-groups (Panel B of Figure 4.23). Though visits regardless of diagnosis also decreased during the post-vaccine period (Panel C of Figure 4.23), the degree by which visits for AOM decreased was larger in magnitude.

Figure 4.23: The figure presents the number of primary care visits among children younger than 20 years of age per calendar-month from 1 January 2005 to 31 December 2015. Children are divided into seven age-groups, listed in the figure legend. Panel A shows the monthly number of visits due to acute otitis media and its complications (AOM). Panels B and C, depict the standardized monthly number of AOM visits (Panel B) and all other visits (Panel C) per age-group. The Y-axis represents the number of standard deviations the observed visits are from from the mean of the entire period for each diagnosis and age-group. The horizontal dotted lines represent values that are zero standard deviations from the mean and the vertical dotted lines represent the beginning of the vaccine intervention. Locally estimated scatter-plot smoothing (LOESS) is used to produce an average trend. Panels B and C suggest that the number of both AOM visits and all other visits have decreased in the post-vaccine period, and that AOM visits have decreased to a larger degree.

The posterior predictions of the component models are shown in Figure 4.24. Each posterior prediction is based on the median of the corresponding marginal posterior predictive distribution. The ITS model with offset consistently predicted the fewest visits in the post-vaccine period. The ITS model without offset consistently predicted the the highest number of visits.

Figure 4.24: The observed and predicted number of visits for acute otitis media and its complications (AOM) from 1 January 2005 to 31 December 2015 for each age-group. Observed visits are illustrated as black points and the predicted number of visits are drawn as lines for each of the component models. The start of the vaccine period is delineated with a vertical black dotted line. Each component model was fitted to the observed visits in the pre-vaccine period, and then used to predict the number of visits in the post-vaccine period, had the vaccine not been introduced. The distance between the observed and predicted visits for each calendar-month is depicted with a thin black line. Longer distances suggest a larger discrepancy. Note that the scale of the Y-axis differ between age-groups.

These component models were stacked using LOOCV to produce the final stacked model. The weights used to stack the component models are shown in Table 4.23.

Table 4.23: The weights used to produce the final stacked model from the component models are presented. The weights for each component model were obtained by minimizing the leave-one-out mean squared error.

Disease category	Age-group	Synthetic controls	ITS with offset	ITS without offset	STL + PCA
AOM visits	0y	0.221	0.000	0.121	0.659
AOM visits	1y	0.149	0.000	0.610	0.241
AOM visits	2y	0.000	0.000	0.479	0.521
AOM visits	3-4y	0.661	0.000	0.339	0.000
AOM visits	5-9y	0.726	0.000	0.274	0.000
AOM visits	10-14y	1.000	0.000	0.000	0.000
AOM visits	15-19y	0.018	0.000	0.078	0.904
Pneumonia hospitalizations	0-4y	0.912	0.001	0.087	0.000
Pneumonia hospitalizations	5-19y	1.000	0.000	0.000	0.000
Pneumonia hospitalizations	20-39y	0.246	0.124	0.000	0.629
Pneumonia hospitalizations	40-64y	0.241	0.000	0.000	0.759
Pneumonia hospitalizations	65-79y	0.000	0.934	0.066	0.000
Pneumonia hospitalizations	80+	0.000	0.472	0.528	0.000
IPD hospitalizations	0-4y	0.001	0.999	0.000	0.000
IPD hospitalizations	5-64y	1.000	0.000	0.000	0.000
IPD hospitalizations	65y+	1.000	0.000	0.000	0.000

The posterior predicted AOM visits and 95% credible intervals are are shown in Figure 4.25. With few exceptions, the observed number of AOM visits are fewer than predicted in the post-vaccine period, indicating that the vaccine prevented visits from occurring.

Figure 4.25: The observed and predicted number of AOM visits from 1 January 2005 to 31 December 2015 for each age-group. Observed visits are illustrated as black points, the posterior predicted visits are presented as lines and 95% credible intervals as a shaded area. The start of the vaccine period is delineated with a vertical black dotted line. The distance between the observed and predicted visits for each calendar-month is depicted with a thin black line. Assuming that the model is correct and that no intervention had taken place, the black points would have an equal probability of occurring above and below the prediction line. Points below the lower bound of the shaded area would then represent observations that would have had less than a 2.5% probability of occurring. Given that the majority of points are located below the prediction line, and many located below the lower bound of the shaded area, the figure suggests that the vaccine resulted in fewer AOM visits. Note that the scale of the Y-axis differ between age-groups.

The rate ratios between the observed and predicted number of AOM cases are shown in Table 4.24. The 95% credible interval of the rate ratio was lower than one in all age-groups, indicating that there was a 97.5% or greater probability that the rate of AOM decreased due to the introduction of PHiD-CV10 in all age-groups. The decrease was largest among young children; 16% (12%-36%) in children younger than one year of age and 18% (5%-42%) in children one year of age. A 12-month rolling rate ratio between the observed and predicted number of AOM cases is presented in Panel A of Figure 4.26. Visually, the rate of AOM cases among children younger than one seems to begin to decline in January 2012, and cases among children one year of age seems to decline in July 2012.

Table 4.24: The rate ratio between observed and predicted number of primary care visits due acute otitis media and complications (AOM) during the post-vaccine period (2013-2015), is presented with 95% credible intervals (95% CI) for the seven age-groups included in the study. The predicted cumulative number of prevented cases as of 1 December 2015 is also presented. A negative number indicates that there is a non-zero probability that the vaccine caused more AOM visits to occur. Direct and indirect savings are presented in 2015 USD.

Age-group	Rate ratio (95% CI)	Cumulative cases prevented (95% CI)	Direct savings (95% CI)	Indirect savings (95% CI)
0y	0.74 (0.64-0.88)	3,234 (1,008 to 5,195)	305,330$ (90,933$ to 514,848$)	45,386$ (11,143$ to 84,654$)
1y	0.72 (0.58-0.95)	5,802 (817 to 11,526)	530,468$ (57,564$ to 1,150,759$)	74,298$ (3,778$ to 193,180$)
2y	0.88 (0.66-0.98)	900 (-185 to 3,817)	92,117$ (-52,649$ to 407,227$)	14,377$ (-11,004$ to 64,562$)
3-4y	0.86 (0.69-0.97)	1,702 (21 to 3,576)	135,274$ (-16,985$ to 357,905$)	23,880$ (-4,324$ to 62,811$)
5-9y	0.88 (0.73-0.96)	979 (229 to 2,521)	134,548$ (-38,612$ to 430,729$)	14,242$ (-1,030$ to 40,961$)
10-14y	0.83 (0.75-0.92)	720 (411 to 1,086)	113,333$ (4,669$ to 285,816$)	10,313$ (-3,098$ to 20,035$)
15-19y	0.89 (0.56-0.98)	430 (210 to 1,689)	55,819$ (-8,278$ to 227,493$	6,169$ (698$ to 25,248$)

The cumulative number of prevented AOM cases reflect both the rate of AOM cases in each age-group, and the consistency and magnitude of the vaccine effect. The cumulative prevented cases per age-group as of December 2015 are presented Table 4.24. The largest effects are seen in the youngest age-groups, who both had the highest baseline rates and experienced the largest relative declines following vaccine introduction. The cumulative number of prevented cases as a function of time during the post-vaccine period is shown in Panel B of Figure 4.26.

Figure 4.26: The impact of the 10-valent Haemophilus influenzae protein D pneumococcal conjugate vaccine (PHiD-CV10) on acute otitis media and complications (AOM) among children younger than 20 years of age is summarized. In Panel A, the estimated 12-month rolling rate ratio between observed and predicted AOM cases is shown per age-group, and the 95% credible intervals (CI) are illustrated as a shaded area. Panel B depicts the cumulative number of prevented AOM cases during the post-vaccine period (2011-2015) for each age-group, along with 95% CI. The total cumulative prevented AOM cases regardless of age-group is shown in Panel C.

The total cost of introducing PHiD-CV10 into the Icelandic pediatric vaccination program from 1 January 2011 to 31 December 2015 was 2,652,364$ in constant 2015 USD. The vaccination resulted in 13,829 (7,337 to 21,114) prevented cases of AOM among children younger than 20 years of age by 1 December 2015 (Panel C of Figure 4.26). Given the observed distribution of costs associated with each AOM visit, the direct savings resulting from vaccine-prevented cases was 1,389,900$ (95% credible interval 704,319$ to 2,201,925$). If the vaccine was assumed to have no other benefits other than preventing AOM, and only direct costs were considered, the incremental cost-effectiveness ratio was 91$ (95% credible interval 21$ to 259$) per prevented AOM case from the health care perspective. The vaccine introduction prevented 10,911 days of work lost (95% credible interval 5,116 to 18,801), which translated to 194,152$ (95% credible interval 78,200$ to 364,155$) in productivity gains. The ICER from the societal perspective was 76$ (95% credible interval 6$ to 244$) per prevented AOM case, assuming the vaccine did not result in benefits in other manifestations of pneumococcal infections. When cost-savings due to reductions in hospital admissions for pneumonia and invasive pneumococcal disease were also included, the direct cost of the PHiD-CV10 introduction was -7,463,176$ (95% credible intervals -16,159,551$ to -582,135$) as of 31 December 2015. From the health care perspective, the vaccination program was already cost-saving 7,463,176$ in the first five years of the program. The corresponding ICER was -543$ (95% credible interval -1,508$ to -48$) per prevented AOM case. When days of work lost due to hospitalized pneumonia and IPD cases were also included, the total cost of including PHiD-CV10 in the pediatric vaccination program was -8,164,894$ (95% credible interval -17,197,959$ to -1,004,553$) as of 31 December 2015. The corresponding ICER was -594$ (95% credible interval -1,597$ to -76$) per AOM case prevented.

4.7.2 Population impact on pneumonia hospitalizations

From 1 January 2005 to 31 December 2017, 13,373 hospitalizations for pneumonia were recorded. Monthly pneumonia hospitalizations displayed complex trends over the study period (Panel B of Figure 4.27). Pneumonia hospitalizations increased fairly rapidly during the pre-vaccine period among adults 40 years and older, and subsequently decreased at variable times in the post-vaccine period. Similarly, hospitalizations regardless of diagnosis increased among adults 20 years and older during the pre-vaccine period (Panel C of Figure 4.27).

Figure 4.27: The figure presents the monthly number of hospital admissions for pneumonia and hospitalizations regardless of diagnosis from 1 January 2005 to 31 December 2017. Panel A shows the monthly number of pneumonia hospitalizations. Panels B and C depict the standardized monthly number of pneumonia hospitalizations (Panel B) and all other hospitaliations (Panel C) per age-group. The Y-axis shows how many standard deviations from the mean the observed hospitalizations are by diagnosis and age-group. The horizontal dotted lines represents values that are zero standard deviations from the mean and the vertical dotted lines represent the start of the vaccine intervention. Locally estimated scatter-plot smoothing (LOESS) is used to produce an average trend.

The posterior predictions of the component models are shown in Figure 4.28. The predictions made by the ITS model without offset diverged from the other models for all age-groups older than 20 years of age, and consistently predicted higher numbers of pneumonia hospitalizations.

Figure 4.28: The observed and predicted number of pneumonia hospitalizations from 1 January 2005 to 31 December 2017 for each age-group. Observed cases are illustrated as black points and the predicted number of cases are drawn as lines for each of the component models. The start of the vaccine period is delineated with a vertical black dotted line. Each component model was fitted to the observed number of cases in the pre-vaccine period. They were then used to predict the number of cases that would have occurred in the post-vaccine period, had the vaccine not been introduced. The distance between the observed and predicted cases for each calendar-month is depicted with a thin black line. Longer distances suggest a larger discrepancy between observed and predicted cases. Note that the scale of the Y-axis differ between age-groups.

These component models were stacked using LOOCV to produce the final stacked model. The weights used to stack the component models are shown in Table 4.23. The predicted number of cases and 95% credible intervals are are shown in Figure 4.29. During most of the post-vaccine period, the observed number of hospitalizations were equal to or below the prediction line among children zero to four years of age, and among adults 20 to 39, 65-79 and 80 years of age and older.

Figure 4.29: The observed and predicted number of pneumonia hospitalizations from 1 January 2005 to 31 December 2017 for each age-group. Observed cases are illustrated as black points. The predicted number of hospitalizations are presented as lines and 95% credible intervals as a shaded area. The start of the vaccine period is delineated with a vertical black dotted line. The distance between the observed and predicted cases for each calendar-month is depicted with a thin black line. Assuming that the model is correct and that no intervention had occurred, the black points would have an equal probability of appearing above and below the prediction line. Given that the majority of points are located below the prediction line, the figure suggests that the vaccine resulted in fewer pneumonia hospitalizations. Note that the scale of the Y-axis differ between age-groups.

The rate ratios between the observed and predicted number of pneumonia hospitalizations are shown in Table 4.25. Among children zero to four years of age, the posterior median of the rate ratio was 0.67, and the 2.5% credidble limit was 0.51. This was consistent with a 2.5% probability that the rate ratio was lower than 0.51 and a 47.5% probability that the rate ratio layed between 0.51 and 0.67. Though the 97.5% credible limit was above the threshold value of one, there was a 94% probability that the rate ratio was lower than one, and a 90% probability that it was lower than 0.83. Simlarly the posterior median of the rate ratio was 0.74 among children five to 19 years of age, and there was a 90% probability that the rate ratio was lower than one. Among adults 65 to 79 years of age, and 80 years of age and older, the posterior median of the rate ratio was 0.75 and 0.76 respectively, and both had a 97% probability of being lower than one.

A 12-month rolling rate ratio between the observed and predicted number pneumonia hospitalizations is presented in Panel A of Figure 4.30. Visually, the rate of pneumonia hospitalizations among children zero to four years of age seems to begin to decline in January 2012 (the first rolling 12-month period to include only post-vaccine months), and hospitalizations among adults 65 years of age and older seems to begin to decline in January 2014.

Table 4.25: The posterior median of the rate ratio between observed and predicted number pneumonia hospitalizations during the post-vaccine period (2013-2017) is presented with 95% credible intervals (95% CI) for the six age-groups included in the study. The predicted cumulative number of prevented cases as of 1 December 2017 is also presented. A negative number indicates that there is a non-zero probability that the vaccine caused more pneumonia hospitalizations to occur. Direct and indirect savings are presented in 2015 USD.

Age-group	Rate ratio (95% CI)	Cumulative cases prevented (95% CI)	Direct savings (95% CI)	Indirect savings (95% CI)
0-4y	0.67 (0.51-1.39)	142 (-115 to 307)	444,533$ (-44,181$ to 1,309,917$)	52,535$ (-59,043$ to 136,715$)
5-19y	0.74 (0.54-1.35)	52 (-27 to 113)	234,848$ (-236,236$ to 748,522$)	20,472$ (-18,876$ to 61,481$)
20-39y	0.68 (0.51-0.95)	182 (14 to 384)	968,662$ (-203,048$ to 2,567,059$)	70,071$ (-9,442$ to 164,747$)
40-64y	0.92 (0.79-1.22)	141 (-270 to 445)	933,290$ (-2,748,49$ to 4,848,557$)	71,953$ (-113,414$ to 223,171$)
65-79y	0.75 (0.55-1.02)	666 (-49 to 1,648)	5,476,585$ (-910,021$ to 15,590,280$)	323,964$ (-4,745$ to 786,252$)
80+	0.76 (0.56-1.02)	631 (-76 to 1,615)	4,664,256$ (-817,266$ to 13,013,699$)	287,270$ (-37,961$ to 742,168$)

The cumulative prevented pneumonia hospitalizations per age-group as of December 2017 are presented in Table 4.25. The largest effects were seen in adults 65 years of age and older, which reflects the baseline number of cases. The predicted cumulative number of prevented hospitalizations as a function of time during the post-vaccine period is shown in Panel B of Figure 4.30.

Figure 4.30: The population impact of the 10-valent Haemophilus influenzae protein D pneumococcal conjugate vaccine (PHiD-CV10) on pneumonia hospitalizations is summarized. In Panel A, the estimated 12-month rolling rate ratio between observed and predicted pneumonia hospitalizations is shown per age-group, and the 95% credible intervals (CI) are illustrated as a shaded area. Panel B depicts the cumulative number of prevented pneumonia hospitalizations during the post-vaccine period (2011-2017) for each age-group along with 95% CI. The total cumulative prevented pneumonia hospitalizations regardless of age-group is shown in Panel C.

The total cost of introducing PHiD-CV10 into the Icelandic pediatric vaccination program from 1 January 2011 to 31 December 2017 was 3,451,805$ at constant 2015 USD. In total, the introduction of PHiD-CV10 resulted in 1,844 (589 to 3,239) prevented pneumonia hospitalizations in the Icelandic population by 1 December 2017 (Panel C of Figure 4.30). Given the observed distribution of costs associated with each pneumonia hospitalization, the direct savings resulting from vaccine-prevented hospitalizations was 13,330,902$ (95% credible interval 2,933,955$ to 26,270,332$), in constant 2015 USD. If the vaccine is assumed to have no other benefits than preventing pneumonia hospitalizations, and only the direct costs are considered, the ICER was -5,315$ (95% credible interval -8,877$ to 711$) per prevented pneumonia hospitalization, indicating a net savings of 5,315$ for each prevented hospitalization from the health care perspective. The vaccination program prevented 29,969 days of work lost (95% credible interval 9,964 to 52,900), which translated to 838,952$ (95% credible interval 273,559$ to 1,493,478$) in productivity gains. From the societal perspective, the ICER was -5,794$ (95% credible interval -9,275$ to 24$) per prevented pneumonia hospitalization, assuming no other vaccine benefit, which implies that the society gains 5,794$ in constant 2015 USD for every pneumonia hospitalization prevented by investing in PHiD-CV10. If the vaccination program’s effects on the other manifestations of pneumococcal disease were included, then the ICER was -5,640$ (95% credible interval -10,336$ to -1,032$) in constant 2015 USD from the health care perspective as of 31 December 2015. Addtionally including loss of work resulted in an ICER of -7,440$ (95% credible interval -13,701$ to -1,175$).

4.7.3 Population impact on hospital admissions for invasive pneumococcal disease

From 1 January 2005 to 31 December 2016, 338 hospitalizations for culture confirmed invasive pneumococcal disease were recorded. Of those, 206 occurred before the introduction of PHiD-CV10 into the pediatric vaccination program in Iceland. Hospital admissions due to vaccine-type IPD were 175, of which 138 occurred prior to vaccine introduction. Only two vaccine-type IPD hospitalizations of children zero to four years of age were recorded in the post-vaccine period. Both cases were unvaccinated and both occurred in 2011. This is compared to 32 hospital admissions of the same age-group in the pre-vaccine period. The number of vaccine-type IPD cases were not sufficiently many to perform a time series analysis.

Standardized hospitalizations for IPD decreased among children zero to four years of age, while standardized hospital admissions regardless of cause did not decrease to the same extent (Panels B and C of Figure @ref(fig:figure-results-paper6-ipd_arranged)). Discrepancies between hospital admissions for IPD and all-cause hospitalizations were also noted in the other age-groups. Hospitalizations for IPD among individuals five to 64 years of age decreased while all-cause hospitalizations remained stable. While hospital admissions for IPD among adults 65 years of age and older did not change visibly, the standardized all-cause hospitalizations increased, suggesting a relative decline in IPD admissions.

Figure 4.31: The figure presents the number of hospitalizations per year-quarter from 1 January 2005 to 31 December 2016. The population is divided into three age-groups, listed in the figure legend. Panel A shows the absolute quarterly number of hospital admissions due to invasive pneumococcal disease (IPD) regardless of serotype. Panels B and C, depict the standardized quarterly number of IPD hospitalizations (Panel B) and all-cause hospitalizations (Panel C) per age-group. The Y-axis represents the number of standard deviations from the mean hospitalizations for each quarter and each age-group. The horizontal dotted lines represent values that are zero standard deviations from the mean and the vertical dotted lines represent the start of the vaccine intervention. Locally estimated scatter-plot smoothing (LOESS) is used to produce an average trend. Panels B and C have been magnified to emphasize the interpretation of the trend line. Panels B and C show that standardized hospitalizations for IPD decreased in all age-groups, relative to the standardized hospitalizations regardless of cause.

The posterior predictions of the component models are shown in Figure 4.32. Both the ITS models consistently predicted fewer IPD cases among children zero to four years of age in the post-vaccine period, compared to the STL + PCA and synthetic control models.

Figure 4.32: The observed and predicted number of IPD hospitalizations from 1 January 2005 to 31 December 2016 for each age-group. Observed cases are illustrated as black points and the predicted number of cases are drawn as lines for each of the component models. The start of the vaccine period is delineated with a vertical black dotted line. Each component model was fitted to the observed number of cases in the pre-vaccine period. They were then used to predict the number of cases that would have occurred in the post-vaccine period, had the vaccine not been introduced. The distance between the observed and predicted cases for each year-quarter is depicted with a thin black line. Longer distances suggest a larger discrepancy between observed and predicted cases.

The stacked model for children zero to four years of age was comprised of the synthetic control model weighted at 0.001 and ITS with offset weighted at 0.999. For individuals five to 64 years of age, and adults 65 years of age and older, the LOOCV procedure assigned full weight to the synthetic control model, excluding contributions from the other three.

The posterior prediction of IPD hospitalizations and 95% credible intervals are are shown in Figure 4.33. Among children zero to four years of age, observed IPD hospitalizations were equal to or fewer than the predicted hospitalizations in all but two quarters. Similary, observed hospitalizations among indvidiuals five to 64 years of age were fewer than predicted more often than expected. Both suggest that the vaccine prevented cases from occurring.

Figure 4.33: The observed and predicted number of IPD hospitalizations from 1 January 2005 to 31 December 2016 for each age-group. Observed cases are illustrated as black points, and the predicted number of cases are presented as lines with 95% credible intervals as a shaded area. The start of the vaccine period is delineated with a vertical black dotted line. The distance between the observed and predicted cases for each year-quarter is depicted with a thin black line. Assuming that the model is correct and no intervention had taken place, the black points would have an equal probability of occurring above and below the prediction line. Given that the majority of points are located below the prediction line, the figure suggests that the vaccine resulted in fewer IPD hospitalizations.

The rate ratios between the observed and predicted number of IPD hospitalizations in the post-vaccine period are shown in Table 4.26. The posterior median of the rate ratio for children younger than five years of age was 0.27, corresponding to a 50% probability that the vaccine impact was greater than or equal to 73%. The 95% credible intervals of the rate ratio were wide, reflecting the uncertainty due to the few number of IPD hospitalizations. However, 90% of the MCMC draws of the rate ratio were below 0.75 and 93% were under the threshold value of one. The 95% credible interval of the rate ratio among individuals five to 64 years of age was lower than one, indicating a 97.5% or greater probability that the rate of IPD hospitalization decreased in this age-group following the introduction of PHiD-CV10.

The 12-month rolling rate ratio is presented in Panel A of Figure 4.34. The rolling rate ratio for children zero to four years of age was unstable due to numerical issues with both the numerator and the denominator. In some 12-month periods, no IPD hospitalizations were observed and the resulting rate ratio was zero regardless of the denominator. In other periods, 2.5% or more of the MCMC draws predicted zero IPD hospitalizations, which resulted in a 95% credible intervals of the rate ratio that extended towards infinity. These issues do not change the overall interpretation of the prediction line presented in Panel A of Figure 4.34 or the rate ratios presented in 4.24.

Table 4.26: The rate ratio between observed and predicted number of hospital admissions for invasive pneumococcal disease (IPD) during the post-vaccine period (2013-2016) is presented along with 95% credible intervals (95% CI) for the three age-groups. The predicted cumulative number of prevented cases as of 1 December 2016 is also presented. A negative number indicates that there is a non-zero probability that the vaccine caused more IPD hospitalizations to occur. Direct and indirect savings are presented in 2015 USD.

Age-group	Rate ratio (95% CI)	Cumulative cases prevented (95% CI)	Direct savings (95% CI)	Indirect savings (95% CI)
0-4y	0.27 (0.05-3.00)	14 (-2 to 67)	227,087$ (71,363$ to 618,919$)	16,882$ (6,893$ to 38,718$)
5-64y	0.44 (0.31-0.68)	29 (1 to 65)	321,424$ (-455,573$ to 1,649,171$)	12,983$ (-3,606$ to 33,498$)
65y+	0.94 (0.62-1.53)	10 (-16 to 45)	73,395$ (-256,856$ to 516,864$)	4,340$ (-10,903$ to 23,543$)

The cumulative prevented IPD hospitalizations per age-group as of December 2016 are presented Table 4.26, and are shown as a function of time in Panel B of Figure 4.34. The posterior median of the cumulative prevented cases increases from the beginning of the post-vaccine period among children zero to four years of age.

Figure 4.34: The population impact of the 10-valent Haemophilus influenzae protein D pneumococcal conjugate vaccine (PHiD-CV10) on hospital admissions for invasive pneumococcal disease is summarized. In Panel A, the estimated 12-month rolling rate ratio between the observed and predicted number of IPD hospitalizations in the post-vaccine period (2011-2016) is shown per age-group. Panel B depicts the cumulative number of prevented IPD hospitalizations during the post-vaccine period (2011-2015) for each age-group along with 95% credible intervals. The total cumulative prevented IPD hospitalizations regardless of age-group is shown in Panel C.

In total, by December 2016, the introduction of PHiD-CV10 prevented 57 (10 to 127) cases of IPD serious enough to warrant hospital admission (Panel C of Figure 4.34). The total cost of introducing PHiD-CV10 into the Icelandic pediatric vaccination program from 1 January 2011 to 31 December 2016 was 3,097,861$, in constant 2015 USD. Given the observed distribution of costs associated with each IPD hospitalization, the direct savings resulting from vaccine-prevented hospitalizations of IPD was 673,008$ (95% credible intervals -189,654$ to 2,081,594$). If the vaccine was assumed to have no other benefits than preventing IPD hospitalizations, and only the direct costs are considered, the ICER was 30,134$ (95% credible interval 8,488$ to 80,375$) per prevented IPD hospitalization. The vaccine introduction prevented 1,280 days of work lost (444 to 2,410) due to IPD, which translated to 35,280$ (95% credible intervals 9,437$ to 70,609$) in productivity gains. If the vaccine was assumed to have no other benefits than preventing IPD hospitalizations, the ICER from the societal perspective was 30,134$ (95% credible intervals 8,487$ to 80,375$) per prevented IPD hospitalization. When cost-savings due to reductions in AOM visits or hospital admissions for pneumonia were also included, the ICER was -119,992$ (95% credible interval -387,183$ to -9,542$) per prevented IPD hospitalization from the health care perspective. When days of work lost were also considered, the ICER was -130,791$ (95% credible interval -416,004$ to -15,860$) per prevented IPD hospitalization.

References

Youngster, Ilan, Jerry Avorn, Valeria Belleudi, Anna Cantarutti, Javier Díez-Domingo, Ursula Kirchmayer, Byung-Joo Park, et al. 2017. “Antibiotic Use in Children – A Cross-National Analysis of 6 Countries.” The Journal of Pediatrics 182 (March): 239–244.e1. https://doi.org/10.1016/j.jpeds.2016.11.027.