IHP 515 Module 6 Causation Paper Example

Did the Clinic Closure Cause the Deaths? Causal Inference in an Overdose Crisis

[Student Name]

Southern New Hampshire University

IHP 515: Population-Based Epidemiology

Module Six Paper

[Instructor Name]

[Date]

Did the Clinic Closure Cause the Deaths? Causal Inference in an Overdose Crisis

In April 2020, Harlan County's only opioid treatment program closed for four months after a staffing shortage during the pandemic. Overdose deaths in the second and third quarters of 2020 were 64% higher than in the same months of 2019. At a public meeting, a commissioner said the closure had killed people and demanded that the provider be held accountable. The association is real, but is it causal? This paper applies epidemiologic tools for causal inference to the question and explains what the evidence can and cannot support.

From Association to Causation

Hill (1965) offered a set of considerations, not rules, for judging whether an observed association is likely causal. They include the strength of the association, its consistency across settings and studies, specificity, temporality, meaning the cause precedes the effect, a biological gradient or dose-response pattern, plausibility, coherence with existing knowledge, experimental evidence and analogy. He stressed that none is required or sufficient on its own, and that the aim is to decide whether there is any other way of explaining the facts that is more likely than cause and effect.

Applying Hill's Considerations

Temporality is satisfied: the closure preceded the rise. Plausibility and coherence are strong, since losing methadone and buprenorphine access is known to increase overdose risk, as cohort studies of people leaving treatment have shown. There is a weak dose-response signal: deaths among the program's former patients were highest in the months when the program was fully closed and fell somewhat after partial reopening. However, strength and specificity are limited. Deaths rose among people who had never attended the program as well as among its former patients, and neighboring counties without closures also saw increases in 2020, which weakens consistency as evidence for the closure specifically.

Multiple Causes Acting Together

Rothman and Greenland (2005) describe the sufficient-component cause model, in which a disease or death results from a combination of component causes that together form a sufficient cause, and many different combinations can produce the same outcome. A single component cause may be necessary in some combinations but not others, and the strength of any one factor depends on how common the other components are. They also argued that causal inference cannot be reduced to checklists and must rest on judgment informed by evidence and alternative explanations.

Applied to Harlan, an overdose death might require a person with opioid use disorder, exposure to fentanyl-contaminated drugs, using alone and lack of treatment. The closure removed treatment for about 300 patients, completing a sufficient cause for some of them. But fentanyl's rapid spread in 2020 and pandemic-related isolation, which increased solitary drug use, were component causes acting across the whole population.

Chance, Bias and Confounding

Grimes and Schulz (2002) explain that before concluding an association is causal, epidemiologists must consider chance, bias and confounding. Chance is a concern with small numbers; Harlan's quarterly deaths number in the teens, so some of the 64% increase could reflect random variation. Information bias could arise if toxicology testing or death certificate completeness changed during the pandemic. Confounding is the largest issue: the closure occurred at the same time as fentanyl's spread and pandemic disruptions, both of which increased overdose risk independently. Because these factors coincided, the crude association overstates the closure's own effect.

What Additional Analysis Could Show

Several analyses could clarify the closure's role. Comparing death rates among the program's former patients with those among similar people who were never enrolled would address confounding by population-wide factors. Comparing Harlan's change with that of similar neighboring counties without closures, a difference-in-differences approach, would help separate the closure from regional trends. Examining whether deaths among former patients rose more sharply than among others would test specificity.

Table 1. Hill's Considerations Applied

Note. Assessments are qualitative judgments based on composite data.

Estimating the Closure's Share

A rough estimate can put the claim in proportion. Before the closure, the program served about 300 patients. Suppose, drawing on cohort studies of people who stop medication treatment, that losing treatment doubled their annual overdose death risk from about 1% to 2% during the months of closure. For four months, that would add roughly one additional death among former patients, perhaps two if risk rose more sharply. Harlan's excess deaths in the second and third quarters of 2020, compared with 2019, numbered about nine. By this crude estimate, the closure might account for a small share of the excess, with the remainder attributable to fentanyl, isolation and other factors. The estimate rests on assumptions and should be replaced by direct analysis of former patients' outcomes, but it shows why a single-cause explanation is unlikely.

A Qualified Conclusion

The evidence supports a qualified conclusion: the clinic closure was probably a contributing cause of some overdose deaths in 2020, particularly among the program's former patients, but it was not the sole or main cause of the county-wide increase, which was driven largely by fentanyl and pandemic conditions. This conclusion matters for policy. Blaming one provider would miss the larger causes and could discourage providers from operating in the county, while ignoring the closure would miss the lesson that treatment access must be protected during emergencies.

Communicating Causal Uncertainty

The epidemiologist's report to the board should explain that the closure likely contributed to some deaths, that several other causes were at work, and what additional analyses could show. Plain language, a simple chart of deaths among former patients versus others and a clear recommendation, such as a plan to keep treatment available during future emergencies, will serve decision makers better than either an accusation or a dismissal.

Conclusion

Causal inference requires weighing evidence rather than accepting the most obvious story. Hill's considerations, the component cause model and attention to chance, bias and confounding show that the clinic closure was one piece of a larger causal picture, a conclusion that points to protecting treatment access while addressing the drug supply and isolation that drove most deaths.

References

Grimes, D. A., & Schulz, K. F. (2002). Bias and causal associations in observational research. The Lancet, 359(9302), 248-252. https://doi.org/10.1016/S0140-6736(02)07451-2

Hill, A. B. (1965). The environment and disease: Association or causation? Proceedings of the Royal Society of Medicine, 58(5), 295-300. https://doi.org/10.1177/003591576505800503

Rothman, K. J., & Greenland, S. (2005). Causation and causal inference in epidemiology. American Journal of Public Health, 95(Suppl. 1), S144-S150. https://doi.org/10.2105/AJPH.2004.059204