Study design, instruments, and analytical framework for investigating determinants of water security program effectiveness in northern Kenya
This page synthesizes the research methodology from the IRB submission documents for the "Community Water Systems" study, a descriptive cross-sectional quantitative study supported by qualitative methods. The study investigates how institutional, financial, and technological factors determine the effectiveness of water security programs across five arid and semi-arid (ASAL) counties in northern Kenya.
The study employs a descriptive cross-sectional quantitative design supported by historical data analysis (2017–2026) to provide longitudinal context. Data collection spans two drought seasons in 2026 across Turkana, Isiolo, Wajir, Garissa, and Marsabit counties.
Percentage of operational days in a given reference period, measured through Virridy sensor data (continuous monitoring since 2017) and direct observation checklists.
Standardized 12-item survey instrument capturing access, reliability, adequacy, and emotional stress dimensions of water insecurity. Scored 0–36.
E. coli and total coliform counts (CFU/100 mL) at point of collection and point of use, using Aquagenix compartment bag test kits.
Virridy datasets from 2017–2026: run-time logs, repair details, spare parts purchased, costs, logistics, and preventive maintenance records.
| Level | Instrument | Method | Sample | Duration |
|---|---|---|---|---|
| Household | Structured survey + HWISE + DCE | Quantitative | 300 households | 25 min |
| Borehole Operator | Semi-structured KII | Qualitative | Each selected borehole | 15–20 min |
| Community Users | Focus group discussion | Qualitative | 6–10 per group, gender-segregated | 20–25 min |
| Water Committee | Focus group discussion | Qualitative | 6–10 WMC members per site | 20–25 min |
| Institutional | Semi-structured KII | Qualitative | County WASH officers, NGO staff | 15–20 min |
| Comparative | Semi-structured KII | Qualitative | ANTHC representatives (Alaska) | — |
| Infrastructure | Observation checklist | Quantitative | Each sampled borehole | — |
| Water Quality | Aquagenix CBT lab testing | Quantitative | All sampled boreholes | — |
| Historical | Secondary data review | Quantitative | 2017–2026 Virridy datasets | N/A |
Sampled boreholes are categorized into three tiers of intervention intensity, enabling dose-response analysis:
Villages are categorized into intervention and non-intervention groups based on DRIP FUNDI program participation, then randomly selected using a random number generator.
Within each selected village, systematic random sampling is applied: the total number of households is divided by the target number of interviews to yield a sampling interval (n). A random starting household is selected, then every nth household is visited.
The primary quantitative instrument is a structured household survey (~25 minutes) administered via the mWater mobile data collection platform on handheld devices. The survey comprises seven substantive sections plus the embedded HWISE scale and discrete choice experiments.
The Household Water Insecurity Experiences (HWISE) scale is a validated, cross-cultural 12-item instrument measuring multiple dimensions of water insecurity. Each item is scored on a 4-point frequency scale (Never = 0, Rarely = 1, Sometimes = 2, Often/Always = 3) with a 1-month recall period. Total scores range from 0 to 36.
The DCE component tests two hypotheses about household preferences for water service management models and willingness to pay, using a simplified paired-choice design.
After the choice task, respondents answer a series of probing questions to understand the reasoning behind their preferences:
Qualitative data is collected through key informant interviews (KIIs) and focus group discussions (FGDs) at multiple institutional levels, providing contextual depth to triangulate with the quantitative findings.
15–20 minute semi-structured interviews covering:
15–20 minute semi-structured interviews covering:
Comprehensive interview guide for the comparative international study site, covering water governance in Indigenous Arctic communities:
20–25 minute group discussions (6–10 participants, gender-segregated):
Note: Facilitators instructed not to mention DRIP FUNDI unless respondents raise it, to reduce social desirability bias.
20–25 minute group discussions (6–10 WMC members):
FGDs organized separately for male, female, and youth groups. Facilitator and note-taker are same gender as participants to accommodate cultural and religious norms.
All instruments administered in local languages: Somali, Borana, Turkana, Rendile, and Samburu by locally recruited enumerators.
Community FGD guide explicitly prohibits mentioning DRIP FUNDI unless respondents raise it, preventing leading questions and social desirability bias.
Quantitative household data triangulated with qualitative FGD/KII data, observational data, water quality lab results, and historical sensor data.
Biological water quality analysis is conducted at each sampled borehole using the Aquagenix most probable number compartment bag test (CBT) kit, measuring E. coli and total coliform concentrations at both point of collection and point of use.
Analysis follows three sequential phases: data preparation and validation, quantitative/inferential analysis, and integration with qualitative interpretation.
Data validation, cleaning, and preparation using the mWater platform export. GPS coordinates replaced with temporary household IDs for de-identification.
Mean, median, standard deviation, IQR for each outcome domain. Disaggregated by county, program participation status, and management model.
Kernel density mapping and Moran's I spatial autocorrelation in ArcGIS/QGIS. Spatial overlays for water quality patterns.
Independent samples t-tests, chi-square tests, one-way ANOVA with Tukey's HSD. Effect sizes reported: Cohen's d, Eta-squared, Cramer's V.
OLS/Beta regression (uptime), OLS/Ordered Logit (HWISE), logistic regression (water quality). 95% CI, p<0.05 significance threshold.
Difference-in-Differences (DiD) using 2017–2026 historical Virridy sensor data to establish counterfactual trends.
Water Security Effectiveness Index (0–1) from z-score normalized uptime, HWISE, and microbial safety with expert-validated weights. Internal consistency via Cronbach's alpha.
Thematic content analysis of FGD and KII data to contextualize quantitative findings. Qualitative insights used to explain statistical patterns and identify mechanisms.
Consent is obtained in person at the beginning of each data collection encounter via the mWater survey form. Verbal consent is provided after an informed consent process conducted in the respondent's local language. No written signature is required. Participation is entirely voluntary with no incentives, penalties, or compensation.
GPS coordinates in raw data could theoretically enable re-identification. Mitigated by de-identification (replacing GPS with temporary household IDs) before analysis.
Mitigated by mWater platform security: HTTPS/TLS encryption in transit, encryption at rest, role-based access control, AWS-compliant cloud hosting.
University of Colorado Boulder, University of British Columbia (Canada), Oslo School of Architecture and Design (Norway), Millennium Water Alliance (Kenya), Urban-A (Norway), RESEAU Centre for Mobilizing Innovation (Canada), Oxfam (South Sudan), Lytton First Nation (Canada). Collaboration involves comparison of results from parallel study sites, not sharing of raw data.
Results shared with DRIP FUNDI program management and disseminated through academic publication. DRIP FUNDI is responsible for sharing water quality results with affected beneficiaries and advising on contamination findings.