NINE-TIPS: A worked example of a prevalence systematic review in mental health

Article type
Authors
Madrid E1, Metzendorf M2, Solà I3, Gabaldà A4, Villahur N5, Munar L5, Serra C6, Rivera F6, Roqué M3
1Interdisciplinary Centre for Health Studies (CIESAL), Cochrane Chile Associate Centre Universidad de Valparaíso Chile. , Viña del Mar, Valparaíso, Chile; Iberoamerican Cochrane Centre - Institut de Recerca Sant Pau (IR SANT PAU) , Barcelona, Spain
2Insitute of General Practice, Medical Faculty of the Heinrich Heine University, Düsseldorf, Germany, Düsseldorf, Germany
3Iberoamerican Cochrane Centre - Institut de Recerca Sant Pau (IR SANT PAU) - CIBERESP, Barcelona, Spain
4Iberoamerican Cochrane Centre - Institut de Recerca Sant Pau (IR SANT PAU) , Barcelona, Spain
5European Agency for Safety and Health at Work (EU-OSHA), Bilbao, Spain
6IMIM-Hospital del Mar Research Institute / University Pompeu Fabra , Barcelona, Spain
Abstract
"Background: The scarcity of methodological guidance for prevalence systematic reviews (PSR) has led to high variability in conduct and report. A ready-to-use collection of tips for authors would be useful.

Objective: To offer tips from a worked example of a PSR on mental health problems in European healthcare workers during COVID-19.

NINE-TIPS framework :

1. FORMULATE A FOCUSED QUESTION:
CoCoPop-S : Condition, context, population and study design

2. DEFINE ELIGIBILITY CRITERIA
Clear definition of condition to be assessed.
Diagnostic criteria (using validated scales/clear cutoff points).
Define subgroups a priori.
Designs to include: cross-sectional, baseline data in cohorts.
Geographical context/population of interest.
Sampling/recruitment process.

3. DEFINE SEARCH CRITERIA AND LIMITS
1. Prevalence: validated filters for prevalence studies not available, consider:
a) synonyms/close terms for prevalence (e.g. frequency, rates, incidence)
b) discarding the concept prevalence/ combining only population (e.g. healthcare workers) + condition (e.g. burnout) + geographic entities (e.g. Europe).
2. Condition (C), population (P):
a) validate terms for C and P by analysing a representative set of studies,
b) use search platforms that permit adjacency operators to create sensitive searches for C and P.
3. Test if strategy identifies known relevant records.

4. SCREENING PROCESS
Two reviewers using multiphase platforms (e.g. Covidence, Rayyan, Distiller).

5. DATA EXTRACTION
Usual: Heterogeneity and different instruments to assess condition.
Align your dichotomisation criteria with articles' criteria.
Extract by duplicate, quality assurance essential.
Check supplementary files.

6. ASSESSING RISK OF BIAS
Many instruments are available, no gold standard.
Consider an instrument that assesses:
a) sample frame
b) type of sample and coverage
c) sample size
d) identification of condition: accuracy and reliability
e) appropriate statistics
f) response rate
(Suggestion: JBI Checklist - may need adaptations).

7. DATA SYNTHESIS
a) Prevalence estimates need transformation before pooling; random effects model advised.
b) Consider subgroup and sensitivity analyses or metaregression, heterogeneity expected.
c) If numerical synthesis not possible, consider following SWiM (Synthesis Without Meta-Analysis) guidelines for narrative synthesis.

8. ASSESSING THE CERTAINTY OF EVIDENCE:
GRADE for PSR unavailable.
Properly designed studies/population representativeness: high initial certainty
No representativeness: lower initial quality

9. REGISTERING PROTOCOL IN PUBLIC REPOSITORY OR PUBLISH"