The impact of health care strikes on patient mortality: A systematic review and meta‐analysis of observational studies

Abstract Objective This study sought to evaluate the impact of health care strike action on patient mortality. Data Sources EMBASE, PubMed CINAHL, BIOETHICSLINE, EconLit, WEB OF SCIENCE, and grey literature were searched up to December 2021. Study Design A systematic review and meta‐analysis were utilized. Data Collection/Extraction Random‐effects meta‐analysis was used to compare mortality rate during strike versus pre‐ or post‐strike, with meta‐regression employed to identify factors that might influence the potential impact of strike action. Studies were included if they were observational studies that examined in‐hospital/clinic or population mortality during a strike period compared with a control period where there was no strike action. Principal Findings Seventeen studies examined mortality: 14 examined in‐hospital mortality and three examined population mortality. In‐hospital studies represented 768,918 admissions and 7191 deaths during strike action and 1,034,437 admissions and 12,676 deaths during control periods. The pooled relative risk (RR) of in‐hospital mortality did not significantly differ during strike action versus non‐strike periods (RR = 0.91, 95% confidence interval 0.63, 1.31, p = 0.598). Meta‐regression also showed that mortality RR was not significantly impacted by country (p = 0.98), profession on strike (p = 0.32 for multiple professions, p = 0.80 for nurses), the duration of the strike (p = 0.26), or whether multiple facilities were on strike (p = 0.55). Only three studies that examined population mortality met the inclusion criteria; therefore, further analysis was not conducted. However, it is noteworthy that none of these studies reported a significant increase in population mortality attributable to strike action. Conclusions Based on the data available, this review did not find any evidence that strike action has any significant impact on in‐hospital patient mortality.

What is known on this topic • Strike action in health care is a contentious issue raising a range of ethical, regulatory, and legal questions.
• While a number of studies suggest strike action has minimal impacts on patient outcomes, this remains disputed.

What this study adds
• This review did not find any evidence that strike action had any impact on in-hospital mortality.
• Country, duration, and the profession on strike had little impact on in-hospital mortality.
• While only three studies examined population mortality, these found that mortality was not attributable to strike action.

| INTRODUCTION
Strike action when carried out by health care workers has been a particularly contentious issue, with debate and controversy spanning several decades. A strike is distinct from other forms of workplace activism or resistance in that it involves a temporary withdrawal of labor as a means to raise some kind of grievance. Such action has been a frequent occurrence across the globe and has varied in length and scale, with strikes lasting hours to hundreds of days, 1 impacting local clinics to entire countries. 2 Perhaps more broadly, the political climate in which strikes have occurred, and the health care systems and patients they have impacted have varied substantially. 3 While demands have most frequently been about workplace pay and conditions, strike action has been utilized to raise a range of other grievances. 4 Strikes when undertaken by health care workers raise a range of ethical, regulatory, and legal issues. One issue that weighs heavily in discussions relates to the well-being of patients, namely the impact that strike action could have on patient care. Arguably the most cited and contentious concern relates to patient mortality. 5 One does not have to look far to find polarizing debate in the literature, with arguments that point out that "[t]he sick and the wounded are regarded as outside the battlefield even in bitter and bloody conflicts." 6 Such concerns have also weighed heavily for professional and regulatory bodies, for example, during the 2016 UK junior doctors strikes, the General Medical Council (GMC; the UK's regulatory body for doctors) issued a stark warning, urging for strike action to be called off, stating that, "we believe that, despite everyone's best efforts, patients will suffer." 7 While these debates are likely to persist, the impact of strikes can and has been measured, with a growing literature examining the impact of strike action on health care delivery, health worker attitudes, and importantly, in this case, patient mortality. Given this and given the fact that strike action will continue to be a frequent feature of health care into the foreseeable future, 8 there is a pressing need for greater clarity in relation to its impact on patient mortality. This study therefore seeks to examine the impact of health care strike action on in-hospital and population mortality.

| REVIEW QUESTION
What is the impact of health care strike action on patient mortality outcomes globally?

| METHODS
A systematic review was carried out to identify all relevant studies related to patient mortality during strike action. Our approach followed PRISMA guidelines. 9 Our study protocol was registered on PROSPERO (registration number CRD42021238879), peer reviewed, and published. 10  In addition, grey literature was searched through SIGMA REPOSITORY.

| Search strategy
While we had planned to search OPEN GREY as per our protocol, it was archived in mid-2021. Search terms were developed to capture the core concepts related to the form of intervention we were interested in (e.g., strike action, industrial action) and the populations in question (e.g., doctors, nurses, health care professionals).
The final search terms were strike OR "industrial action" OR "industrial dispute" OR "collective action" AND doctor OR physician OR clinician OR "medical practitioner" OR nurs* OR "health profession*" OR healthcare OR "health care" OR "pharmac*" OR "dentist" OR "midwi*" OR dieti* OR "occupational therap*" OR "paramed*" OR "physiotherap*" OR "radiograph*" OR "psycholog*" OR "health worker" OR "hospital." There were no publication dates or language restrictions. Where complete data for a relevant outcome was not available, we contacted authors to request data. In addition, we conducted a manual search of reference lists of eligible studies.

| Eligibility
We included observational studies (cross-sectional or cohort studies) that examined in-hospital/clinic or population mortality during a strike period compared with a control period where there was no strike action. We only included papers where health care professionals went on strike (as opposed to non-professional health care staff) and where health care services were directly impacted by the strike, not services that dealt with the "upstream" effects of a strike, for example, a hospital where staff were not on strike, but that dealt with excess patients from a nearby hospital on strike (Table 1).

| Primary outcomes
• In-hospital/clinic mortality during a strike period as a proportion of admissions was examined against a comparable time period before and/or after the strike action. Mortality rates for each period were compared.
• Population mortality during a strike period was examined against a comparable time period before and/or after the strike action. Mortality rates for each period were compared.

| Screening and data extraction
Two authors (RE and SMW) conducted the first screen of titles and abstracts to confirm eligibility. A second screen was then undertaken by RE and SMW, which examined the full text of the remaining articles against the above eligibility criteria. Disagreements were resolved through discussion or with a third member of the review team. For the studies retained, RE extracted study data, which were checked by SMW. Data were extracted related to study characteristics, the nature of the strike, the outcomes of the study, and any other contextual details (Table 2).

| GRADE assessment and quality appraisal
To assess the certainty of the evidence, the GRADE approach 11 was applied. This approach specifies four levels of certainty for a body of evidence for a given outcome: high, moderate, low, and very low.
Studies are assessed against five criteria: risk of bias, imprecision, inconsistency, indirectness, and publication bias. Observational studies start with a "low" rating, and this can be either increased or decreased against the above criteria. Additionally, study quality was assessed using the NIH quality assessment tool 12 for observational cohort and cross-sectional studies and the Cochrane Risk Of Bias In Non-Randomized Studies-of Interventions (ROBINS-I) tool, 13 which rates the potential for study bias arising pre-intervention (confounders, participant selection), during the intervention (classification of intervention), and after intervention (deviations, missing data, outcome measurement, result selection). The quality of the study evidence was evaluated by two authors (EK and RE) and two authors (SD and SMW) examined a random sample of assessments.

| Analysis
Meta-analysis was used to systematically synthesize the findings of the single, independent studies retrieved from the search and included for analysis. The relative risk (RR) was calculated for each study. We pooled RRs using a random-effects model and tested for heterogeneity. RevMan 15 and the metafor package in R 16 were used to carry out the analyses.

| Heterogeneity
We tested for the existence of heterogeneity with Cochran's Q statistic (where p < 0.05 indicates heterogeneity is present). We assessed the magnitude of the variation in effect sizes across studies with Higgin's I 2 statistic, which estimates the proportion of variance in effect sizes due Low to true heterogeneity (from 0% to 100%), with higher values representing greater inconsistency in effect size across studies. We also report τ as a measure of heterogeneity for each comparison, which gives the SD of the effect size estimates. If heterogeneity was high (I 2 > 75%), we planned to conduct a meta-regression to explain possible sources of variation, examining the following potential moderators: strike duration (days); country (low-and middle-income country vs. high-income country); profession on strike (e.g., doctors, nurses, multiple roles), and whether the strike involved a single or multiple facilities. These were chosen as they were identified as being the main differential factors that were consistently reported between studies.

| RESULTS
The

| Risk of bias and study quality
Overall, the certainty of the evidence was "very low" when applying the GRADE approach. Studies mainly had issues on domains related to risk of bias and inconsistency. That is, the majority of studies had at least some limitations in their execution, not including important details about the strike or patient cohorts, heterogeneity was also high. Given the lack of information often found in studies, indirectness was also found to be an issue. The GRADE ratings for each study are included in Table 2, and a GRADE evidence profile is included in Appendix Table 4.  Table 3 for more information.

| Sensitivity analysis
A sensitivity analysis was conducted with Bhuiyan and Machowski 17 removed as this study was of particularly low quality and a clear outlier. As its exclusion had no impact on the overall results, we opted to retain this study.

| Population mortality
While this review also sought to analyze population mortality related to strike action, further analysis was not possible or appropriate. From the above search, five papers were found that contained population mortality data; however, only three were potentially suitable for analysis. Three papers reported data from the same strike that impacted Los Angeles county in 1976, 24,34,35 one reported data from a strike that impacted Jerusalem in 1983 32 and one reported data from a nationwide strike in Croatia in 2003. 20 Given that three studies examined the same strike, it only left three studies that could be included in any potential analyses. While further analysis was not possible, it is perhaps noteworthy that none of the above studies reported a significant increase in population mortality that was attributable to strike action.

| DISCUSSION
Based on data that were available to us, this review did not find any evidence that strike action has a significant impact on in-hospital T A B L E 3 Meta-regression analysis examining whether differences in mortality across strike versus non-strike periods were moderated by length of strike, country economy, profession, and number of facilities on strike patient mortality. Furthermore, the impact of strike action does not appear to be affected by country, duration, or profession on strike.
While we were unable to analyze the impact of strike action on population mortality, the small number of studies that were found did not report any significant increase in population mortality that could be attributed to strike action.
Caution, however, is warranted in interpreting these results.
Firstly, the above studies overall were of a relatively low quality scoring poorly on all three quality appraisal instruments. This means that we can only have minimal confidence in the results reported above.
There could of course also be alternative explanations as to why strike action did not impact mortality that we also cannot rule out; namely, many strike-impacted hospitals were able to put contingency plans in place, minimizing the disruption and impact on patient mortality.
While several studies provided detailed accounts of the contingencies put in place during strike action to minimize disruption, many did not.
Secondly, and related to these points, the nature of the strikes reported varied substantially, as did the health care systems they impacted. In many papers little detail was included about the nature of the strike or the context in which it was occurring, for example, the number of staff on strike or how well resourced the health care system was to cope with the action. While we have included four variables (country, length, staff on strike, and whether the strike occurred in a single or multiple facilities), these should not be seen to capture all nuances and features of a strike. Furthermore, because of the lack of information included in studies, it was often not clear how or if these factors contributed to any disruption, for example, even strikes that we might intuitively expect to cause the most disruptive (for example, those that were protracted and involved multiple facilities), actually did not, with a number of studies suggesting this was actually not the case. Ruiz 29 for example examined a nationwide strike in the United Kingdom where >90% of hospitals functioned as normal.
Thirdly, while this review focused on mortality, it would be insightful to assess the impact of strike action on health-related quality of life and patient satisfaction with care. Fourthly, a number of studies also failed to report patient characteristics during strike and control periods, that is, most studies have little information on those who sought care during a strike versus those who sought care in non-strike periods; we therefore cannot be sure if or how this impacted the above results. Finally, there are several studies examining the upstream impacts of strike action (that were found in this search but subsequently excluded) that suggest that strike action had a significant impact on presentations and mortality. 36 It is possible that while presentations and mortality decreased in strike-impacted hospitals, many sought treatment elsewhere.
Following this, there are several implications for future research.
Firstly, there is a need for future studies that examine strike action to include greater detail about the cohorts being examined and the nature of the strike. There also needs to be a broader examination of factors beyond in-hospital mortality, linking these data with F I G U R E 2 Forest plot of risk-ratios on the impact of strike action on in-hospital mortality population statistics about mortality. We also found few studies that linked mortality and other patient outcomes. Secondly, there is also a need to better understand how strike action changes access to care.
The limited literature that does exist suggests that generally, while many will delay seeking care, others will seek care elsewhere. 35 Future studies will benefit from integrating these insights alongside mortality data to understand how strike action impacts mortality in strikeimpacted facilities and those dealing with its upstream effects.
While our results say little about the impact that strike action has on health care delivery, our results are consistent with a broader body of work that suggests that strike action has minimal impact on a range of other patient outcomes, 37 and past work on the impact of strike action on mortality. 38 At a minimum, this review suggests that strike action by health care workers can be conducted safely as it relates to patient mortality. This has several implications for debates in relation to the impact of strike action, its justification, and the other legal and regulatory concerns that such action raises. Most notably, while patient outcomes are a valid concern that should weigh heavily in discussions about the justifiability of strike action, strike action should not be dismissed on this point alone as it is far from inevitable that patients will be harmed when health care workers go on strike.