Increasing Influenza Vaccination Rates in People With Chronic Illness

Abstract

Background

The safety and efficacy of influenza vaccination for the chronically ill are clearly supported by the evidence, yet vaccination rates in this vulnerable population remain low. This leads to many avoidable hospitalizations and deaths in Germany every year. The goal of this systematic review is to identify measures in primary care medicine that can be used to increase influenza vaccination rates among the chronically ill.

Methods

This review was carried out as recommended in the PRISMA statement. A systematic literature search was performed. Only randomized, controlled trials were included in the analysis. Details can be found in the study protocol (PROSPERO, CRD42018114163).

Results

15 trials were included in the analysis. Training sessions for medical practice teams focusing on a particular disease raised the vaccination rates by as much as 22%. A financial incentive had the greatest effect (relative risk [RR]: 2.79; 95% confidence interval: [1.18; 6.62]). Reminders via text message yielded a maximum 3.8% absolute increase in vaccination rates. Complex interventions were not found to be of any greater benefit than simple ones.

Conclusion

A variety of approaches can be effective. Focusing training sessions for medical practice teams on certain diseases may be of greater benefit than vaccination-centered training sessions. Reminder systems for doctors should be more reliably implemented. Simple strategies are perhaps the most suitable ones in the heterogeneous population of chronically ill persons. The limitations of this systematic review include the heterogeneity of the studies that we examined and the small number of studies in each category.

In the 2017/2018 influenza season, around 45 000 patients were admitted to hospital for treatment of seasonal influenza in Germany alone; 1674 deaths due to this disease were reported to the Robert Koch Institute (e1). The course of the illness and the risk of hospitalization or death vary depending on the general health status of the individual patient. The young, the elderly, and the chronically ill are especially vulnerable (1, e1). For instance, a study published in 2017 showed that 86% of patients admitted to the hospital with an “influenza-like illness” (ILI) had at least one chronic (= 65 years) medical condition. The prognosis was poor despite adequate treatment: more than 30% of the patients needed intensive care, and 14.8% of them died in the hospital (1). Besides the individual consequences, the additional consultations and hospitalizations caused by seasonal influenza represent a heavy burden on the entire healthcare system. Chronically ill patients have frequent contacts with their primary care physicians (e2), who therefore carry a large share of the extra load (2).

The treatment of influenza is primarily symptomatic; in the early stage, the population at risk should also be given antiviral drugs (3). The most effective way of combating the virus is prophylactic vaccination, shown in many studies to be efficacious and safe, particularly with reference to the chronically ill (4, e3– e6).

A vaccination rate of 90% among the chronically ill is necessary to attain herd immunity (5). However, this quota is rarely achieved. Depending on country, population, and age group, rates of between 9% and 70% have been observed (6– 8). In Germany, the vaccination rate in chronically ill patients below 60 years of age was 24.1% in the season 2012/2013 and 22.6% in 2013/2014 (9). Among those over 60, the rate was 43.6% in 2010/2011, but only 34.8% in 2016/2017 (e7). Consequently, it is essential to improve vaccination rates, especially in high-risk populations. Given their coordinating role in the healthcare system and their professional focus, primary care physicians would appear especially well suited to implement these interventions.

The aim of our study was to evaluate measures taken to increase the rates of vaccination against seasonal influenza among people with chronic illness in the primary care setting.

Methods

This systematic review was conducted in accordance with the PRISMA Statement for systematic reviews (e8). A detailed description of the methods can be found in the study protocol (PROSPERO CRD42018114163) and in the eMethods (10). We included randomized controlled trials that had been carried out by or were focused on primary care physicians and which examined strategies to increase the rate of vaccination against influenza in the chronically ill. The systematic literature survey (ebox) was performed in October 2018.

eBOX. Search strategy for MEDLINE and Embase via Ovid.

((chronic* or longterm or condition* or multiple or persist* or ongoing or degenerat* or high risk or disab* or illn* or diseas* or congenital or incurab* or defect* or disorder* or sick* or syndrom* or disposition* or disabil* or failure or asthma or copd or cirrhos* or hepatit* or renal insuff* or diabet* or immunodefic* or immunosuppr* or scleros* or HIV or human immunodeficiency virus or AIDS or Acquired immunodeficiency syndrome or cancer* or malign* or tumor*)

and

(influenza* or flu)

and

(system* or strateg* or program* or campaign* or educat* or participat* or remind* system* or recall* or postcard* or letter* or *phone campaign*/ or *phone call*/ or financial incentiv* or encouragement or motivation or buisness hour* or opening hour* or home visit* or house call* or bureaucracy or transfer* or responsib* or competen* or liabil* or inform* or remind* or commendat*)

and

(immun* or vaccin* or unvaccin* or unimmun* or inoculat*)

and

(rate or uptake or coverage or target or percentage)

and

(randomized controlled trial or controlled clinical trial or randomized or placebo or drug therapy or randomly or trial or groups or double blind or rct or controlled)).tw.

Thesaurus search strategy for MEDLINE via PubMed

Search,Query,Items found,Time

#6,“Search (#1 and #2 and #3 and #4 and #5) Sort by: [pubsolr12]“,4,04:05:51

#5,“Search clinical trials, randomized[MeSH Terms]“,121371,04:04:55

#4,“Search (immunization[MeSH Terms]) OR vaccination[MeSH Terms]“,164265,04:04:43

#3,“Search ((((preventive health service[MeSH Terms]) OR home care[MeSH Terms]) OR medical staff[MeSH Terms]) OR health behavior[MeSH Terms]) OR health, attitude to[MeSH Terms]“,961187,04:04:18

#2,“Search influenza[MeSH Terms]“,45235,04:03:35

#1,“Search (((((((((chronic disease[MeSH Terms]) OR congenital disorder[MeSH Terms]) OR asthma[MeSH Terms]) OR chronic obstructive lung disease[MeSH Terms]) OR cardiomyopathies[MeSH Terms]) OR immunosuppression[MeSH Terms]) OR diabetes mellitus[MeSH Terms]) OR cancer[MeSH Major Topic]) OR human immunodeficiency virus[MeSH Terms]) OR acquired immune deficiency syndrome[MeSH Terms]“,4783699,04:03:20

The Cochrane risk-of-bias tool was used to assess the risk of bias in the studies included for analysis (e9). After assigning the individual criteria 0 points for high, 1 point for unclear, and 3 points for low risk of bias, we summed the risk scores to form a quality factor (QF; modified according to [11]), which served to enhance comparability. On the basis of this factor the quality of each trial was rated as low (0–10 points), moderate (11–14 points), or high (at least 15 points).

We preferentially extracted adequately corrected results (e10) from cluster-randomized studies. Whenever possible, intention-to-treat analyses were carried out. Owing to the high methodological, clinical, and statistical heterogeneity of the trials, meta-analysis with calculation of a pooled effect estimator was not a suitable means of synthesizing our data.

Results

Fifteen trials were included for analysis (figure 1) (12– 26). The populations studied varied widely. At least 77% of the patients included were below 65 years of age. Eleven of the trials had investigated patients with cardiovascular disease, while the remaining four focused on chronic obstructive pulmonary disease (COPD) or asthma. There were no studies on patients with mental illness. The categorization of the trials is shown in Table 1, their characteristics in Table 2 and the eTable. Besides their methodological and clinical heterogeneity, the trials exhibited marked statistical heterogeneity: assessed according to Higgins and Thompson (e11), the heterogeneity of the studies was I²= 87 % for those focusing on training of office teams, I²= 85% for those on enhancing the competence of medical professionals, and I²= 94% for those on sending text-message reminders.

Figure 1.

Flow chart for selection of studies

Table 1. Categorization of the interventions.

Focus on medical personnel (n = 7)			Focus on patients (n = 8)
Classified according to:			Classified according to:
EPOC*1	Complexity		Form of reminder	Additional features
Training programs for office teams (n = 3) Reminder systems for physicians (n = 2) Extending competence of medical professionals*2 (n = 2)	Simple*3 (n = 5) Complex*4 (n = 2)		Text message (n = 2) Postcard (n = 2) Letter (n = 3) Educational brochure (n = 1) Financial incentive (n = 1)	Personalized*5 (n = 4) Educational element*6 (n = 4)

*¹ Cochrane Effective Practice and Organization of Care Review Group (EPOC) (e26), taxonomy of the different intervention categories in healthcare systems

*² Extension of the spectrum of competence of medical professionals, e.g., physician assistants

*³ No more than two different intervention components

*⁴ At least three different intervention components

*⁵ Patient reminders with personal message or individual recommendation by patient’s primary care physician

*⁶ Patient reminders with informative elements on the symptoms, risks, and side effects of influenza and vaccination against influenza

Table 2. Characteristics of the studies included for analysis.

Author, year	Design	QF	No. of patients*1	Population	Control	Interventions
Focus on medical personnel							Comp.
Beck 1997 (14)	RCT	9	321	≥ 65 years	No intervention	Modification of professional roles: group visit model	3+
Chambers 1991 (15)	cRCT	14	203	<65 years	No intervention	Reminder system for physicians: On documentation forms 1) for all patients 2) for half of the eligible patients	1
Hermiz 2002 (16)	RCT	12	177	30–80 years, COPD	No intervention	Modification of professional roles: follow-up home visits by community nurses after hospital discharge	3+
Markun 2018 (18)	cRCT	18	216	≥ 45 years, COPD	No intervention	Training program for office teams: COPD care package	2
Siriwardena 2002 (23)	cRCT	12	SG 1: 6207 SG 2: 4327 SG 3:   169	No restrictions	Usual care + feedback	Training program for office teams: Provision of information and recommendations for action + feedback	2
Tierney 2005 (25)	cRCT	15	363	≥ 18 years, bronchial asthma or COPD	No intervention	Reminder system for physicians: with regard to important preventive treatments, at computer workplaces	1
Zwar 2016 (26)	cRCT	16	256	40–85 years, COPD	Usual care + diagnostic training + copy of guidelines	Training program for office teams: COPD care package + diagnostic training + copy of guidelines	2
Focus on patients							Feature(s)*2
Ahmed 2004 (12)	cRCT	12	2067	18–64 years	−	Postcard: 1) once only; 2) twice at a 1-month interval	−
Baker 1998 (13)	RCT	10	14 004	1) ≥ 65 years 2) <65 years	No intervention	Postcard/letter: 1) standard postcard 2) personalized postcard 3) personalized letter	1) E 2) PE 3) PE
Herrett 2016 (17)	cRCT	18	102 257	18–64 years	No intervention	Text message	−
Moran 1992 (19)	RCT	10	207	<65 years	No intervention	Letter: 1) once only 2) twice at a 1-month interval	PE
Moran 1996 (20)	RCT	13	234	<65 years	No intervention	Educational brochure and/or financial incentive: 1) educational brochure 2) chance to win food vouchers 3) combination	1) E
Mullooly 1987 (21)	RCT	7	2217	≥ 65 years	No intervention	Letter	PE
Regan 2017 (22)	RCT	17	6245	≥ 6 months	No intervention	Text message	−
Spaulding 1991 (24)	RCT	16	1068	No restrictions	No intervention	Postcard	P

*¹ Total number of patients studied

*² Additional features: personalized (P), educational element (E)

Comp., number of components in intervention; COPD, chronic obstructive pulmonary disease; cRCT, cluster-randomized controlled trial; QF, quality factor; RCT, randomized controlled trial; SG, subgroup

eTable. Detailed characteristics of the studies included for analysis.

Author Year Country	Design*1	DatA*2	RB*3	Setting*4	No. of patients*5	Patients	Control	Interventions*6	QF*7	Comp. *8
Interventions focusing on medical personnel
Beck 1997 USA (14)	RCT	op	Yes	HMO	C: 161 I: 160 T: 321	– Age ≥ 65 years and – chronic heart, lung, or joint disease or DM combined with – high use of healthcare services	No intervention	Modification of professional roles: group visit model; care of eligible patients during monthly group appointments with other patients in the office, involving educational events and preventive measures by physician assistants, socialization, and, if needed, individual consultations with the primary care physician	9	3+
Chambers 1991 USA (15)	cRCT physician T: 32	reqrd	No	University general medical center	T: 203	– Age < 65 years and –  diagnosis of DM, kidney injury, anemia, chronic heart failure, bronchial asthma or COPD	No intervention	Reminder system for physicians: computer-generated reminder notices for influenza vaccination on documentation forms for 1) all eligible patients or 2) half of the eligible patients	14	1
Hermiz 2002 Australia (16)	RCT	op	Yes	Mixed	C: 93 I: 84 T: 177	– Age 30–80 years and –  status post hospital discharge after treatment of a condition related to COPD	No intervention	Modification of professional roles: patients visited at home by community nurses after hospital discharge; detailed appraisal of condition and circumstances, performance of care required; report to and coordination with respective patient’s primary care physician	12	3+
Markun 2018 Switzerland (18)	cRCT physician T: 33 strat	userd	No	Primary care physician offices	C: 115 I: 101 T: 216	– Age ≥ 45 years and – nicotine abuse, at least 10 pack years – new diagnosis of COPD by spirometry	No intervention	Training events for office teams: Care package for COPD, including educational events for office teams focused on detection and care of COPD	18	2
Siriwardena 2002 UK (23)	cRCT office T: 30 strat	opadj	No	Primary care physician offices	C, SG1: 3182SG2: 2268 SG3: 107 I, SG1: 3025 SG2: 2059 SG3: 62 T, SG1: 6207 SG2: 4327 SG3: 169	–  No age restriction and –  diagnosis of 1) CHD 2) DM 3) previous splenectomy	Usual care + monitoring and feedback system	Training events for office teams: training carried out by study team during group meetings of office team; provision of evidence-based data and simple recommendations for action regarding influenza vaccination + monitoring and feedback system	12	2
Tierney 2005 USA (25)	cRCT physician T: 274	oprd	Yes	University primary care office	C: 169 I: 194 T: 363	– Age ≥ 18 years and – diagnosis of bronchial asthma or COPD	No intervention	Reminder system for physicians: automated reminders regarding important preventive treatments on physicians’ computer systems	15	1
Zwar 2016 Australia (26)	cRCT office T: 36 strat	oprd	No	Primary care physician offices	C: 110 I: 144 T: 256	– Age 40–85 years –  documented nicotine abuse and –  new diagnosis of COPD by spirometry	Usual care + COPD- diagnosis training for physician assistants + copy of guidelines for physicians	Training events for office teams: Care package for COPD, training events for office team regarding team-based disease management + COPD diagnosis training for physician assistants + copy of guidelines for physicians	16	2
Interventions focusing on patients
Ahmed 2004 USA (12)	cRCT employer T: 505 strat	opadj	No	MCO	I1: 929 I2: 1138 T: 2067	– Age 18–64 years and –  chronic cardiovascular, pulmonary, or renal disease, immunosuppression, DM, or hemoglobinopathy	–	Postcards: 1) one reminder postcard 2) two postcards at a 1-month interval	12	–
Baker 1998 USA (13)	RCT	req	No	MCO members and self-funders	SG 1: 10 573 SG 2: 3431 T: 14 004	1) Age ≥ 65 years or 2) age < 65 years and diagnosis of dm, bronchial asthma, nephrotic syndrome, tkf, chd, or sickle-cell anemia	No intervention	Postcards and letters: 1) standard postcard 2) personalized postcard 3) personalized letter	10	1) E, 2/3) PE
Herrett 2016 UK (17)	cRCT office T: 156 strat	opadj	No	Primary care physician offices	C: 51 136 I: 51 121 T: 102  257	– Age 18–64 years and –  chronic cardiac, neurological, respiratory, renal, or hepatic disease or immunosuppression	No intervention	Text messages: Text-message reminders	18	–
Moran 1992 USA (19)	RCT	op	No	Community health center	C: 68 I1: 69 I2: 70 T: 207	– Age < 65 years and –  chronic cardiac, pulmonary, renal, or metabolic disease or hemoglobinopathy	No intervention	Letters: 1) one reminder letter 2) two reminder letters	10	PE
Moran 1996 USA (20)	RCT	op	Yes	Municipal community health center	C: 64 I1: 59 I2: 65 I3: 46 T: 234	– Age < 65 years and –  diagnosis of chronic cardiopulmonary, or metabolic disease, hemoglobinopathy, immunosuppression or cKI	No intervention	Educational brochure and financial incentive: 1) educational brochure sent 2) chance to win food vouchers 3) combination of 1 and 2	13	1) E
Mullooly 1987 USA (21)	RCT	op	No	HMO	C: 1112 I: 1105 T: 2217	– Age ≥ 65 years –  status post discharge from inpatient hospital treatment –  for chronic cardiovascular, pulmonary, renal, metabolic, neurological or malignant disease	No intervention	Letters: personalized letter	7	PE
Regan 2017 Australia (22)	RCT	op	No	Primary care physician offices	C: 3138 I: 3107 T: 6245	– Age ≥ 6 months –  diagnosis of severe bronchial asthma or immune disorder or chronic pulmonary or cardiac disease	No intervention	Text messages: Text-message reminders	17	–
Spaulding 1991 USA (24)	RCT	op	No	Military primary care physician office	C: 549 I: 519 T: 1068	– No age restriction with –  ischemic heart disease, nicotine abuse, pulmonary emphysema, bronchiectases, tuberculosis, nephrotic syndrome, sickle-cell anemia, cKI, cardiac valve disease, chronic bronchitis, COPD, bronchial asthma, mucoviscidosis, pernicious anemia, DM	No intervention	Postcards: reminder postcard	16	P

*¹ Study design in cluster randomization: level of randomization, total number of clusters (T), stratification (strat) carried out during randomization

*² Data used for the review: only published (op), adjusted for randomization mechanism (adj), raw data (rd); additional data requested from author (req), additional data used (use)

*³ Further calculations required for use of data in review

*⁴ Setting as given in source study

*⁵ Numbers of patients in control groups, intervention groups, and subgroups (C, I, SG) and total numbers (T)

*⁶ Category and description of intervention(s) implemented

*⁷ Calculated quality factor for study concerned

*⁸ Number of components in intervention

CHD, Coronary heart disease; cKI, chronic kidney injury; Comp., components; COPD, chronic obstructive pulmonary disease; cRCT, cluster-randomized controlled trial; DM, diabetes mellitus; E, educational component; HMO, health maintenance organization; MCO, managed care organization; P, personalization; RB, review board; tKF, terminal kidney failure

The quality of the trials included for analysis also varied widely (for details see the eFigure). Blinding of outcome assessment was mostly assessed as bearing a low risk of bias, because many authors used objective recording methods. Assessment of the risk of incomplete outcome data or selective outcome reporting was often not possible because study protocols were not available.

eFigure.

Assessment of the risk of bias

Effectiveness of interventions

Patient-centered studies tended to have larger samples. Overall, studies focused on medical personnel were of higher quality (figure 2).

Figure 2.

*¹ Data for individual subgroups, if amalgamation not possible

*² Vaccination rate (VR) in control group (VR C) and absolute difference in VR between intervention group and control group (? VR)

*³ Additional data from authors used

*⁴ Statistically adjusted data used in event of cluster randomization

*⁵ Further calculations carried out on basis of available data

CI, Confidence interval; M-H, Mantel–Haenszel; RR, relative risk; bold type indicates statistically significant results

Interventions focussed on medical personnel

Training for office teams was examined in three trials. COPD care packages, investigated by Markun et al. (18) and Zwar et al. (26), demonstrated significant positive effects (relative risk [RR] 1.35, 95% confidence interval [1.14; 1.59] and RR 1.29 [1.03; 1.62], respectively), with control group vaccination rates of 61.6% and 49.1%. Siriwardena et al. (23) achieved improvement of the vaccination rate by 22% in patients post splenectomy by means of a vaccination-centered workshop, but the sample was very small (n = 169 patients), so the effect size was not significant (RR 1.08 [0.90; 1.25]). Only slight, non-significant improvements were achieved in patients with coronary heart disease or diabetes mellitus (RR 1.02 [1.00; 1.04]); however, very high proportions of patients were vaccinated even without intervention (72.5 % and 70.2 %, respectively). Studies in this category were of moderate to high quality (QF 12 to 18).

Automatic reminder systems for physicians with printed vaccination recommendations on documentation forms proved effective in the trial by Chambers et al. (15) (RR 1.86; p = 0.001). The vaccination rate was lower among participants whose physicians received reminders for only half of their patients. For unknown reasons, digital reminders were also generated for only half of the patients included in the trial by Tierney et al. (25), leading to a vaccination rate 2.2% lower than in the control group (RR 0.95; [0.67; 1.35]). In both cases, the control group vaccination rate was around 20%. Studies in this category were of moderate to high quality (QF 14 to 15).

Programs to enhance the competence of medical professionals were generally complex interventions of low to moderate quality (QF 9 to 12). Beck at al. (14) achieved a significantly positive result by treating patients in the framework of structured group visits under the supervision of medical professionals (RR 1.27 [1.11; 1.46]). In contrast, a study by Hermiz et al. (16) showed that home visits by a community nurse in contact with the responsible primary care physician resulted in a vaccination rate 7.4% lower than in the control group (RR 0.89 [0.70; 1.12]); there were problems with the quality of implementation and the communication between nurse and physician. In both cases, the control group vaccination rate was around 65%.

Altogether, simple interventions achieved greater effects and were more likely to produce significant results than complex interventions.

Interventions focussed on patients

The effect of sending text-message reminders to patients was investigated in two high-quality studies (QF 17 to 18) by Herrett et al. (17) and Regan et al. (22). The former had a vaccination rate of 50% in the control group and were able to achieve an absolute increase of only 1.7% (RR 1.05 [1.00; 1.11]), while the latter attained a significant absolute increase of 3.8% over the rate of 9.1% in the control group (RR 1.41 [1.22; 1.63]).

Personalized postcard reminders were used by Spaulding et al. (24) in a high-quality study (QF 16) and by Baker et al. (13) in a low-quality study (QF 10). Compared with the 9% figure in the control group, Spaulding et al.’s personalized postcard raised the vaccination rate by 16.1% (RR 2.77 [2.05; 3.75]). Baker et. al. found that only a personalized postcard achieved a significant result in both subgroups, with generally greater effect sizes (RR 1.09 and 1.11; p <0.05) than for a non-personalized postcard (RR 1.05 and 1.07; p >0.05). The control group vaccination rate was 35.8 for patients under 65 and 50.7% for those over 65.

Studies in which the patients received reminder letters were of low quality (QF 7 to 10). The letters were personalized and featured an educational element. Only in their younger subpopulation did Baker et al. (13) achieve a significant effect (RR 1.09; p <0.05). In the study by Moran et al. (19), a single reminder letter reduced the vaccination rate by 0.5% in comparison with the control group (RR 0.99 [0.60; 1.63]). Mullooly et al. (21) found a significant increase of 8.8% in the vaccination rate compared with 30.1% in the control group (RR 1.29 [1.15; 1.45]).

Moran et al. (20) investigated the effectiveness of sending an educational brochure and providing a financial incentive, alone and in combination, in a study of moderate quality (QF 13). All three interventions achieved statistically significant improvements over the vaccination rate of 9.4% in the control group. The combination of brochure and financial incentive (RR 2.78 [1.13; 6.87]) was superior to the brochure alone (RR 2.53 [1.04; 6.15]), but not to the financial incentive alone (RR 2.79 [1.18; 6.62]).

Ahmed et al. (12) and Moran et al. (19) compared sending two reminders to each patient with sending only one reminder. In the former study, the vaccination rate fell by 5.8% in 18- to 49-year-olds and rose by 4.6% in 50- to 64-year-olds (RR 0.94 [0.84; 1.05] and RR 1.08 [1.00; 1.15], respectively). In the latter, the rate went down by 6.1% (RR 0.80 [0.46; 1.38]).

Discussion

The identified interventions designed to increase the vaccination rates for seasonal influenza among people with chronic illness in the primary care setting are heterogeneous. We found that both interventions focused on patients and interventions focussed on medical personnel have been implemented successfully.

Interventions focussed on medical personnel

Both Kovacs et al. (11) and Forsetlund et al. (27) conjectured that training programs alone can exert only limited influence on complex behavior patterns. Among the studies we analyzed, training courses for office teams that were oriented on the management of a particular disease (18, 26) achieved better results than a vaccination-centered approach (23). Consequently, training programs with an obvious bearing on routine practice may be effective; however, the effect should not be overestimated.

Reminder systems for physicians were effective only when the messages were generated for all eligible patients (15, 25). Therefore, any such system needs to be highly reliable. A related approach involves standardized checklists: these are intended to help assess the indication for vaccination and were found by Mendu et al. (28) and Merkel et al. (29) to have a positive effect on vaccination rates.

Group visits to enhance the competence of medical professionals showed positive effects (14), whereas cooperation between community nurses and primary care physicians may be difficult (16). A related approach is the blanket delegation of vaccination—indications and implementation—by physicians to their office nurses (30). Overall, there is broad evidence for the efficacy of extending the competence of medical professionals (31); indeed, this is now common practice in countries such as the UK and the USA (32).

Interventions focussed on patients

Text-message reminders to remind patients about vaccination, investigated in two recent high-quality studies (17, 22), showed no advantage over letters or postcards. Both studies found that such reminders achieved only slight increases in vaccination rates. The great strength of this type of intervention is its wide scope; the two studies analyzed here had extremely large sample sizes (102 257 and 6245 patients respectively). Consequently, there is sufficient evidence for the positive effect of text-message reminders in increasing chronically ill patients’ adherence to their physicians’ recommendations (e12).

Overall, the implementation of patient reminder systems in Germany seems inadequate (33); universal coverage would be desirable. Written reminders addressed directly to patients can also be effective in Germany, as shown by Schulte et al. (34) in a sample of patients with chronic kidney disease.

A number of publications described the use of financial incentives as an element of influenza vaccination campaigns (e13). Like other researchers before us (20), we found that this type of intervention had the greatest effect.

A personal recommendation by the patient’s own doctor has been described as a powerful predictor of subsequent vaccination (e14, e15). The credibility of the source is also a major factor in the assessment of the indications for vaccination and the associated risks following an educational program (e16). Personalized invitations to attend for screening colonoscopy showed a positive effect in Germany (35, 36). Our findings do not permit any clear conclusions with regard to the respective effects of personalized and non-personalized invitations, but overall one can by all means assume a benefit of personalization.

Educational interventions have previously been described as an appropriate way of filling gaps in knowledge (e17) and improving vaccination rates (e18). Addition of an educational element to patient reminders yielded indifferent results in our study, however, leading to neither an increase nor a decrease in vaccination rates.

Complexity of interventions

In the studies we investigated, simple interventions displayed higher quality and wider-reaching effects than complex interventions. A second reminder had no additional effect, and the inclusion of an educational element did not necessarily improve the end result. The combination of an educational brochure and a material incentive was not significantly superior to either measure alone. All this supports the previous observation that complex interventions do not automatically generate greater effects (11, 31, e19). Simple approaches may therefore prove eminently suitable means of improving the vaccination rate in large populations with heterogeneous attitudes and requirements.

Perspective

Authors have described promising interventions to increase the vaccination rates of at-risk populations in settings other than primary care and outpatient clinics. One suitable environment may be the emergency room, where the proportion of patients at risk is much higher than in the total population (37). Rimple et al. (38) increased the influenza vaccination rate of high-risk patients in a surgical emergency department from 16% to 83% by screening all patients for indications. Screening before certain operations or in the presence of particular clinical constellations in the inpatient setting, as reported by Gurfinkel et al. (39), also led to higher vaccination rates. As outlined by Schulte et al. (34), vaccination programs in Germany could in future be implemented largely by health insurance funds or federal institutions.

Limitations

Although our search covered the four major relevant databases and the gray literature, pertinent publications may have been overlooked. The quality of the studies we evaluated was variable, and some had very small samples. Since we found no significantly negative results, publication bias cannot be ruled out. The Pearson correlation coefficient was significantly negative for the control group vaccination rate and the RR per study (r = -0:650; p = 0.01); consequently, our results could also suffer from ceiling effects (40). In some cases we had to use the results of cluster-randomized studies without sufficient correction, thus impairing quality and power. The studies were highly heterogeneous in terms of patients, country, healthcare system, and interventions, which particularly limits the external validity of our findings.

Supplementary Material

eMethods

Methods

This systematic review was conducted in accordance with the PRISMA Statement for systematic reviews (e8). A detailed description of the methods can be found in the study protocol (PROSPERO CRD42018114163) (10).

Inclusion criteria and literature survey

We adjusted the definition of the World Health Organization (WHO) (e20) to define chronic illness as any long-lasting, slowly progressing, communicable or non-communicable, physical or mental disorder. The only eligible setting was general primary care (American Association of Family Physicians; AAFP [e21]) in industrialized countries (United Nations Statistics Division; UNSD [e22]). We included randomized controlled trials that had been carried out by or were focused on primary care physicians and which examined strategies to increase the rates of influenza vaccination among the chronically ill.

We performed a keyword and thesaurus search of titles and abstracts in the databases Embase, MEDLINE, CENTRAL, and ERIC for the period from 1946 to October 2018. Moreover, we scrutinized reference lists and registries. Details of the search strategy can be found in the eBox.

Study selection and literature survey

The eligibility of abstracts and full texts was assessed by two authors (FB, LS) working independently. If differences of opinion could not be resolved by discussion, a third author (JG) decided whether or not the study concerned should be included. Data that conformed with the CONSORT 2010 Statement (e23, e24) were documented in a Microsoft Access^® database; outcome data were also registered with Microsoft Excel^® and RevMan (Version 5.3, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Any pertinent data missing from the studies included for analysis were requested from the corresponding authors.

The reported influenza vaccination rate was the sole endpoint documented. Because of the seasonal nature of influenza outbreaks, no restrictions were placed on the duration of follow-up.

Categorization of the interventions

Patient reminders with a personal message, addressed to the patient by name, or containing an individualized recommendation from the patient’s primary care physician were classified as “personalized.” Approaches that conveyed information on the symptoms, risks, and side effects of influenza and vaccination against influenza were classed as interventions with an “educational element.”

Assessment of the risk of bias

The risk of bias was assessed using the Cochrane risk-of-bias tool. After assessment of the individual criteria, assigning 0 points for high, 1 point for unclear, and 3 points for low risk of bias, we calculated a quality factor (QF) for each study analyzed by summing the scores for all types of bias (modifying the method according to Kovacs et al. [11]) in order to improve comparability. The QF values were used to classify the studies into low quality (0 to 10 points), moderate quality (11 to 14 points), and high quality (at least 15 points).

Effect size and statistical methods

Relative risks (RR) and confidence intervals (CI) were calculated using RevMan and Microsoft Excel^®. Other statistical analyses were performed with SPSS^®. Adequately corrected results were preferentially extracted from cluster-randomized trials (e10, e25). Results of comparisons other than intervention versus usual care were marked separately. All analyses were conducted on the intention-to-treat data. If a publication contained two or more eligible subgroups that could not be amalgamated, the subgroups were all included and displayed individually.

The results of the studies we analyzed were not amenable to meta-analysis on grounds of the high heterogeneity of the samples. The results were therefore displayed in forest plots, and the synthesis was narrative. Conclusions as to the effectiveness of the various categories of intervention were drawn based on the proportion of significant positive results.

Key Messages.

• Fewer than half of the chronically ill people in Germany are vaccinated against influenza.

• Primary care is a suitable setting for implementation of interventions to improve the vaccination rate.

• Training programs for office teams that are focused on particular chronic diseases may be more effective than vaccination-centered approaches.

• Automatic patient reminder systems are advisable, but have not yet been installed widely in Germany.

• Reminder systems for physicians must be highly reliable.