Elisha Abraham
Elisha Abraham

Healthcare and research data analyst

I turn health and business data into answers people can act on, and I show my working.

I come from nursing, where I am in the final year of my degree, and I learned analysis by doing it: survey research in SPSS first, then SQL, Python, Power BI and Excel. Every project below starts from raw data anyone can obtain, ends in a plain answer, and can be re-run by anyone from the code in its repository.

Open to remote data analyst roles and freelance analysis projects.

The question comes first

Each project opens with one question and answers it in the first paragraph, with the numbers that matter and what I would do about them.

The numbers are checked twice

Results are recomputed independently, usually in a second tool. One project matches on 1,348 figures, another on 1,297.

The limits are stated

Every write-up says what the data cannot show, so nobody acts on more than the evidence supports.

Projects

Column chart: 30-day readmission rate rises from 8.6% for patients with no inpatient stay in the previous year to 37.1% for those with five or more.

SQL · MySQL 8

Hospital readmissions among patients with diabetes

Which patients come back within 30 days, and where should follow-up effort go first?

101,766 hospital stays, from raw file to audited data model to a risk score tested on patients it was not built on. 4.5% of patients account for half of all 30-day readmissions, and a five-item score sorts discharges into tiers from 6% to 27%.

Bar chart with confidence intervals: share of adults who went without medical care several times or more, from 73.8% in the North East to 35.7% in the South West.

SPSS · Complex Samples

Going without medical care in Nigeria

Who goes without care, and what happens to those who reach a public clinic?

A national survey of 1,600 adults analysed for its sample design. The public file has no cluster numbers, so I rebuilt and verified them. 57.5% went without care; it follows poverty and place, not sex or age.

Chart of the share of US adults by glycaemic status, including diagnosed and undiagnosed diabetes.

Python · survey statistics

Undiagnosed diabetes in US adults

How many adults have diabetes without knowing it, and which screening rule finds them?

NHANES 2021 to 2023 with design-based statistics written from first principles and checked against statsmodels. 2.3% of adults, 5.5 million people, are undiagnosed; the standard screening rule reaches 68% of them.

Histogram of p-values from 4,850 A/B tests: 45% fall below 0.05, far more than chance alone would give.

Python · experimentation

What 27,000 headline experiments say about A/B testing

How often is there a real winner, and how much of a win is luck?

Findings were made on one part of the data, written down, then tested once on the rest. A winner measured at +30% is worth about +22%, and all seven predicted headline patterns held on new tests.

Line chart of monthly revenue over two years, peaking each November at about 1.4 million pounds.

SQL · MySQL 8

Customer retention and value at a UK online retailer

Do customers come back, and which ones does the business depend on?

1,067,371 invoice lines through cohorts, time to second order and RFM segments. The top tenth of customers bring in 63.2% of revenue, and 69.7% of new customers order again within a year.

Power BI report page comparing Nigerian states on delivery in a health facility.

Power BI · DAX

Maternal and child health across Nigeria's states

Where are mothers and children best and worst served?

Four rounds of Demographic and Health Surveys in one report. 43.3% of births were in a health facility in 2024, from 97.0% in Imo to 8.8% in Kebbi.

Excel dashboard of employee attrition with filters, headline figures, six charts and a table of groups to watch.

Excel · formulas

Employee attrition dashboard

Who is leaving, why, and what is it costing?

An interactive dashboard where every number is a live formula and the workbook tests itself. A quarter of the "attrition" was dismissals, and overtime staff leave at 30.5% against 10.4%.

Research

Research analysis

Undergraduate research assistant in the Department of Nursing Science, University of Ilorin, since 2024. I clean and analyse data for health and nursing research in SPSS: survey and questionnaire studies, from data entry and coding to tests, tables and the written results.

Publication

Anyebe, E. E., Hamzat, A. T., Madu, T. N., Ejodkun, A., Omitogun, E. O., & Abraham, E. (2026). Compassion fatigue and coping strategies among nurses in a tertiary hospital in North-Central Nigeria: a facility-based cross-sectional study. The Nursing Scope, 9(2), 145–164. doi.org/10.59073/thenursingscope92/145-164

Research case studies

Ten studies I analysed for student and faculty researchers in nursing, medicine and public health. In each one the researchers designed the study and collected the data, and I prepared the data, ran the analysis and reported the results. Names and places are withheld, no data or documents are shared, and findings are summarised in words because the full results belong to the researchers.

Select a study to open it.

Corneal ulcers: a clinical records study774 patient records · chi-square tests · SPSS
Question
What causes corneal ulcers, who is affected, and how do patients' eyes fare after treatment?
Data
Clinical records of 774 patients from all six of Nigeria's geopolitical zones, with 151 variables each: demographics, cause, risk factors, other illnesses, sight at presentation and at discharge, and final outcome.
What I did
Prepared and coded the clinical variables, including fields where one patient can have several diagnoses or risk factors. Summarised causes, risk factors and outcomes, compared groups with chi-square tests, and tabulated sight at presentation against sight at discharge. Wrote separate results reports for infective ulcers (497 patients) and ulcers caused by injury (190 patients).
What it showed
The reports set out the mix of causes, the leading risk factors, including injury, self-medication and traditional eye medicines, and how sight changed between presentation and discharge.
Tools
SPSS 27.
What married men know about danger signs in pregnancySurvey of 233 · logistic regression, Mann-Whitney, Spearman · SPSS
Question
How much do married men know about danger signs and complications in pregnancy, what shapes that knowledge, and is it related to their involvement in maternal care?
Data
A cross-sectional survey of 233 married men in a city in North-Central Nigeria, recruited by three-stage cluster sampling, with 123 variables.
What I did
Scored the knowledge and involvement scales and checked their reliability. Tested associations with chi-square and Mann-Whitney tests, fitted binary logistic regression models for adequate knowledge, and examined knowledge against involvement with Spearman correlation.
What it showed
Knowledge was generally poor and limited to the most visible signs. Age and a belief in the value of education predicted adequate knowledge. The link between knowledge and involvement looked like a threshold, not a steady rise.
Tools
SPSS 27 (syntax).
Medication safety among student nursesSurvey of 204 · Kruskal-Wallis, chi-square, Spearman · SPSS
Question
How well do student nurses know and practise safe medication administration, and what is associated with it?
Data
A cross-sectional survey of 204 nursing students in three year groups on clinical placement at a teaching hospital, chosen by proportionate stratified sampling, with 65 variables.
What I did
Cleaned the questionnaire data, built knowledge and practice scores and checked their reliability. Compared year groups with Kruskal-Wallis tests, tested associations with chi-square, and related knowledge, supervision and training to practice with Spearman correlation.
What it showed
Knowledge and self-reported practice were both high, but knowledge was not related to practice. Supervision was the factor most consistently associated with safer practice, and a sizeable minority accepted unsafe shortcuts.
Tools
SPSS 27 (syntax), Excel.
Hygiene and infectious disease in a communitySurvey of 422 · hygiene score and index, Spearman · SPSS
Question
Are personal and household hygiene related to how often residents report common infections?
Data
A cross-sectional survey of 422 adult residents of one community in North-Central Nigeria, chosen by four-stage sampling and interviewed with a structured questionnaire, with 79 variables.
What I did
Cleaned the data, built a personal hygiene score and a household environmental hygiene index, summarised the infections reported over twelve months, and related both hygiene measures to reported infection with Spearman correlation.
What it showed
Personal hygiene was mostly good, household environmental hygiene much less so, and reported infections were common. Better hygiene on both measures went with fewer reported infections, and the environmental measure had the stronger link.
Tools
SPSS 27 (syntax), Excel.
Self-care among patients with type 2 diabetesSurvey of 119 · published scales, multiple linear regression · SPSS
Question
How well do patients manage their diabetes day to day, and what is associated with better self-care?
Data
A cross-sectional survey of 119 patients at a diabetes outpatient clinic in a teaching hospital. The questionnaire combined published scales for self-care activities, self-efficacy, social support, depression and communication with care providers, giving 129 variables.
What I did
Scored each published scale by its own rules, tested associations with chi-square and Spearman correlation, and fitted multiple linear regression models to find which factors held independently.
What it showed
Self-care was poor for most patients despite good knowledge and confidence. Checking blood glucose and exercise were the least practised activities. Knowledge, self-efficacy, social support and communication with providers were each related to self-care, and communication was the one that held in the regression model.
Tools
SPSS 27 (syntax), Excel.
Mothers' acceptance of the malaria vaccineSurvey of 409 · reliability, Spearman · SPSS
Question
Will mothers of young children accept the malaria vaccine, and how do the benefits and risks they perceive relate to acceptance?
Data
A cross-sectional survey of 409 mothers of children under five in one community, chosen by multi-stage sampling, with a questionnaire built on the Health Belief Model and 141 variables.
What I did
Cleaned and coded the data, built scores for perceived benefits, perceived risks and acceptance, checked their reliability, classified each into levels, and tested the study's two hypotheses with Spearman correlation.
What it showed
Most mothers saw high benefits and accepted the vaccine. Acceptance rose strongly with perceived benefits and fell with perceived risks.
Tools
SPSS 27 (syntax), Excel.
Where women choose to give birthSurvey of 375 · Bloom's cut-off, normality tests, Kruskal-Wallis · SPSS
Question
What do women know and think about where to give birth, and what is associated with their views?
Data
A cross-sectional survey of 375 women aged 15 to 49 in one community in North-Central Nigeria, with 77 variables. The data arrived as a raw export of form responses.
What I did
Turned the raw export into a coded SPSS dataset with a written codebook. Built perception and knowledge scores, checked their reliability and classified them with Bloom's cut-off. Tested the scores for normality and, because they were skewed, compared groups with chi-square and Kruskal-Wallis tests.
What it showed
Just over half of the women had their last birth in a health facility. A favourable view of facility delivery and good knowledge were both uncommon, and both rose with education and household income.
Tools
SPSS 27 (syntax), Excel.
Health workers' compliance with immunisation guidelinesSurvey of 115 · Mann-Whitney, Kruskal-Wallis, Spearman · SPSS
Question
How closely do health workers follow immunisation schedules and guidelines, and what is associated with compliance?
Data
A cross-sectional survey of 115 health workers in three primary health centres in one city, with 69 variables.
What I did
Scored the knowledge and compliance scales and checked their reliability. Compared groups with Mann-Whitney U and Kruskal-Wallis tests, and related compliance to knowledge, supervision, equipment and the availability of guidelines with Spearman correlation.
What it showed
Knowledge was high and self-reported compliance very high. Compliance was most strongly related to supportive supervision, then to the availability of guidelines and to knowledge. Because compliance was self-reported, the report recommends checking it against clinic records.
Tools
SPSS 27 (syntax), Excel.
Student nurses' experience of clinical placement (mixed methods)Mixed methods: 205 questionnaires and 10 interviews · thematic content analysis · SPSS
Question
What is clinical placement like for student nurses, and what would improve it?
Data
A convergent mixed-methods study: a questionnaire completed by 205 nursing students in three year groups, and in-depth interviews with ten students.
What I did
Analysed the questionnaire descriptively and the interview transcripts by thematic content analysis. Kept an audit trail that checks each of the 124 quotations used against its transcript. Brought the two strands together objective by objective.
What it showed
Students get substantial hands-on experience and their attitudes to learning are positive, but workload, too few instructors, staff attitudes and equipment shortages limit what they gain. The interviews added emotional strain and worries about patient safety, which the questionnaire did not capture.
Tools
SPSS 27, Excel.
Assessing clinical skills in simulation training (qualitative)Qualitative: 8 interview transcripts · thematic analysis · ATLAS.ti
Question
How do instructors and students experience two structured tools for assessing clinical skills?
Data
The qualitative strand of a mixed-methods study led by a university research team: eight interview transcripts from expert and student raters who had used the tools.
What I did
Coded the transcripts in ATLAS.ti following Braun and Clarke's thematic analysis. Built a codebook of six themes and 17 sub-themes, and a coding matrix that ties each of 64 quotations to its source, participant, theme and the reason for the code, so the team can verify every coding decision.
What it showed
The analysis produced six themes on how useful and workable the tools are in training. The findings are the research team's to publish.
Tools
ATLAS.ti, Excel.

Skills and training

Tools

  • SQL (MySQL 8): window functions, CTEs, data modelling
  • Python: pandas, numpy, statsmodels, matplotlib
  • SPSS: syntax, Complex Samples, scale reliability
  • Power BI: data modelling and DAX
  • Excel: formula-driven models and dashboards

Methods

  • Data auditing and cleaning with a written trail
  • Survey analysis: weights, clusters, design effects
  • Regression, including logistic and Poisson models
  • Experiments: power, false discoveries, the winner's curse
  • Cohorts, retention and customer segments

Training

  • Bachelor of Nursing Science, University of Ilorin (final year)
  • Data Analyst programme, The Data Immersed (2026)
  • Data Analysis training, VextorHub (2025)

Contact

The quickest way to reach me is email: elishaabraham25@gmail.com. I am also on LinkedIn and GitHub.