Using an Outbreak Investigations Model for Quality Improvement Studies

Primary care and community health investigators typically are trained to conduct their studies as if they were federally funded research grants, even if those studies are small unfunded projects. A complete theoretical model is expected along with a large sample size, multiple sites, and sophisticated statistical methods. The case definition should meet precise clinical criteria and the measures should be the most rigorous available. The study design should be a randomized experiment. Compromising on any of these elements weakens the study in the eyes of well-funded investigators.

The aforementioned rigorous criteria effectively discourage many investigators. If they can’t do it right, they wonder, then why do it at all? Let me suggest that another paradigm is more relevant to practice-level studies, such as quality improvement (QI) projects. The Centers for Disease Control and Prevention (CDC) teaches field epidemiologists how to investigate infectious disease outbreaks using a pragmatic yet valid and effective approach that is more flexible and ultimately more useful in field settings than the conventional research model. ¹

Nosocomial infection rates, after all, are quality indicators. Let me suggest that other complications, indeed any adverse event, can be addressed using CDC’s outbreak investigations model. A spike in any quality indicator might be regarded as an “outbreak.” The CDC is highly respected around the world for its ability to investigate and manage outbreaks, so primary care and community health investigators should feel no embarrassment about emulating their methods.

When investigating an outbreak, field epidemiologists first assess whether the incidence exceeds the expected number of cases. Note that a complete theoretical model, capturing all likely risk factors, is not needed at this stage. In fact, even the case definition is not fixed and may be refined as the investigation unfolds. After the case definition is established, cases may be classified as probable, possible, or confirmed. Note the lack of certainty about whether a particular patient qualifies as a case. In QI studies, we employ indicators that are statistically associated with poor quality but may not be proof-positive of poor quality in every case. To use a crude example, hospital readmission may be unavoidable in many cases but a high readmission rate might still be a valid quality indicator. The investigator who insists that every readmission be confirmed as occurring due to clinical errors may stifle the study before it can begin.

The CDC offers this guidance:

Early in an investigation, a loose case definition that includes confirmed, probable, and even possible cases is often used to allow investigators to capture as many cases as possible. Later on, when hypotheses have come into sharper focus, the investigator may tighten the case definition by dropping the “possible” category. ¹

For each possible case, investigators capture geographic location, demographic data, date of onset, and risk factors. This information allows for description of data by person, place, and time. Graphing the outbreak across time, drawing the “epidemic curve,” is a standard step, yet one not always presented in published research reports of quality of indicators.

The number of descriptive variables captured may need to be expanded as the investigation proceeds. Primary care investigators may have some advantages over public health investigators chasing down the causes of community outbreaks. The shoe-leather epidemiologist, who may be going door-to-door in an afflicted neighborhood, does not know at the outset what population is at risk, may be unaware of most of the cases, and also must collect a lot of information via interviews. In contrast, the primary care QI investigator may be able to rapidly capture descriptive information from electronic medical records.

Developing the hypotheses is step 6 in the process and evaluating the hypotheses is step 7. At this point, the investigator who has been trained in the grant-funded research paradigm probably has dropped his or her coffee mug in shock. Hypotheses should be set a priori in that model. In field work, we do not want to abandon the study just because we discover an unexpected cause of the quality problem. Instead, we let the descriptive data shape the initial hypothesis tests.

In QI studies, hypotheses can be tested using analytic epidemiology. The 2 main types of analytic studies are cohort studies and case–control studies. The former compares groups of people who have been exposed to suspected risk factors with groups who have not been exposed. The latter compares people experiencing the adverse event (case patients) with a group of people not experiencing the adverse event (controls). Case–control studies are “look-back” studies since the investigator first groups the cases based on the outcome. A cohort study is ideal in a small well-defined population with a high attack rate. But if the adverse event is rare and we are not sure who was at risk, the case–control approach is more efficient.

The results of significance testing may lead to new hypotheses. After the analytic study is completed, control measures are implemented and results are reported. Infectious disease investigators may interrupt transmission or exposure. Or they may reduce susceptibility via vaccinations. QI investigators also may be able to interrupt the chain of events that leads to the adverse outcome, reduce exposure to the cause, or implement protections for people who may be exposed. To continue with the hospital readmission example, investigators may have learned that polypharmacy is a significant risk factor. ² Medication review at discharge might be suggested to correct this problem, thus interrupting the chain of events.

After the study is completed, it should be reported. Readers will be interested in the methods and the rationale behind suggested control measures. And of course a follow-up study to assess the effectiveness of the prevention strategies will be important.

Applying the outbreak investigations approach to primary care and community health studies stands the conventional paradigm on its ear. The flexibility, pragmatism, and inductiveness of the approach might be described by some as diametrically opposed to “good research methods.” Yet CDC routinely uses the model to quickly and effectively deal with life or death problems. This track record attests to its potential usefulness for primary care and community health studies, whether they are described as practice management, QI, outcomes research, or managerial epidemiology. ³