Why pre-screen epidemiologists before the technical panel
The failure mode in epidemiology is not statistical incompetence, which is rare among trained candidates. It is over-claiming: an association reported in language that implies causation, a confounder unaddressed, a model presented with more confidence than the data supports. Those errors are invisible in a resume and become highly visible when a finding reaches policy or the press. A short screen asks how a candidate talks about a prediction that turned out wrong.
What actually matters when screening Epidemiologist candidates
- 01
Method and rigour
Check command of study design: cohort versus case-control selection, confounding control, sample size calculation, and which regression models they ran in R, SAS, or Stata.
- 02
Real casework
Probe actual investigations: outbreak line lists, case definitions written, contact tracing volumes, NNDSS or REDCap surveillance data cleaned, and the population size covered.
- 03
Interpretation and judgement
Test how they read imperfect data: distinguishing artefact from real signal in incidence trends, handling reporting delays, missingness, and when association does not justify causal claims.
- 04
Reporting and testimony
Assess written and verbal output: peer-reviewed papers, MMWR-style field reports, briefings to health officers or clinicians, and handling press or public questions on risk.
Pre-screening questions to ask Epidemiologist candidates
12 questions grouped by what they test. Ask the same set in every screen and score answers on a consistent scale, or send them as an async video screen and compare answers side by side.
Design supports inference
3 questions01Do you have experience creating study designs? Can you share examples?
Listen forA design matched to the question with the confounding strategy named, and an honest account of what the design cannot establish.
Designs described by name only, or no consideration of confounding in an observational study.
02Which types of epidemiological research are you experienced in?
Listen forSpecific study types with the populations and outcomes named, plus an honest statement of where their experience is thin.
Claims coverage across every study type, or research experience limited to analysing datasets others designed.
03Can you share your experience using biostatistics in analysing epidemiological data?
Listen forMethods they applied themselves with the assumptions understood, plus how they handled missing data rather than dropping it silently.
Analysis performed by a statistician with no engagement, or methods applied without checking their assumptions.
Real investigations
3 questions04What is your experience with disease surveillance systems?
Listen forWorking with real surveillance data including its known biases, such as reporting delays and ascertainment differences between areas.
Surveillance data treated as complete, or no awareness of reporting lags and under-ascertainment.
05Can you describe research of yours that had a significant impact on population health?
Listen forA decision or intervention that followed, with their own contribution separated from the wider team and the evidence stated honestly.
Impact claimed with no decision named, or credit taken for a change driven by other factors.
06Can you describe your experience with data collection, analysis and interpretation?
Listen forInvolvement before the data existed, including case definitions and collection instruments rather than analysis of a supplied dataset.
Only ever received clean datasets, or no input into how cases were defined.
Association and cause
3 questions07Can you discuss a time your predictions were inaccurate and how you handled it?
Listen forThe assumption that broke named specifically, with what they published or told stakeholders and what changed in the next model.
Misses attributed entirely to unforeseeable events, or no forecast of theirs that turned out wrong.
08What is your knowledge of disease transmission and the mechanisms behind it?
Listen forMechanistic understanding used to judge whether an association is plausible, rather than statistics applied with no biological reasoning.
Purely statistical treatment with no view on whether a finding makes biological sense.
09Which computational tools and models do you use in your analyses?
Listen forNamed software used for their own work, with models whose assumptions and limits they can state rather than describe as outputs.
Software named with no analysis behind it, or models used without any sensitivity checking.
Findings in public
3 questions10How do you explain complex findings to people with no background in the field?
Listen forUncertainty expressed in usable terms, with association described as association rather than translated into causal language.
Simplifies by removing uncertainty, or uses causal language for observational findings.
11What experience do you have presenting findings to stakeholders or public health teams?
Listen forReal presentation experience including being challenged, with how they handled a question at the edge of what the data supported.
Defensive about challenge, or willingness to answer beyond what the study could establish.
12How would you help implement prevention strategies based on your findings?
Listen forRealism about what evidence can settle, with an intervention they helped shape and honesty about the implementation constraints.
Recommendations made with no view on feasibility, or evidence presented as deciding a question it cannot.
How to score responses
Score every candidate on the same four criteria immediately after the screen. At this stage you are shortlisting for panel interviews, not making the final call.
Method and rigour
35%5Names specific designs used, justifies confounder adjustment and bias controls, and shows fluency in survival or multivariable regression modelling.
Real casework
25%5Describes named outbreaks or cohorts with dates, case counts, and the specific analytic role they personally held, not team-level generalities.
Interpretation and judgement
25%5Cites a case where they resisted or reversed a premature conclusion, explaining the data limitation and the alternative explanation checked.
Reporting and testimony
15%5Points to published or internal reports plus a briefing where technical uncertainty was conveyed clearly to non-epidemiologist decision-makers.
Over-claiming is the failure mode here, and it is invisible on a resume until a finding reaches policy. A one-way video screen asks about the forecast that missed.
Try it on HirevireScreening FAQ
Process basics
How long should a pre-screening round for an epidemiologist take?
Fifteen minutes across eight to ten questions, answered async. Enough to test design reasoning, hear one real investigation, and establish how they handle communicating uncertainty to a non-specialist audience.
Should the screen include a technical exercise?
Not yet. Ask them to describe a study they designed and the confounding they had to address. A statistical exercise is expensive to set and mark, and it does not surface over-claiming, which is the harder problem to detect.
Evaluating answers
What is the strongest signal when screening an epidemiologist?
How they discuss a forecast that missed. Epidemiologists with applied experience name the assumption that broke and what they changed. Candidates who attribute every miss to unforeseeable events have not examined their own models closely.
How do I check they distinguish association from causation?
Listen to the verbs. Strong candidates say a factor was associated with an outcome and state what would be needed to support a causal claim. Anyone describing observational findings in causal language will do the same in a press release.
Go deeper on this role
Sanat has been hiring since 2012 and watching the recruitment industry change up close ever since, and turned that screening process into Hirevire's video screening platform. LinkedIn
