University of Turku Public Mental Health Master's Thesis Guide: PMH0020, 40 ECTS, Epidemiology, Interventions and UTUGradu

Quick answer: what is the University of Turku Public Mental Health thesis route?

The current University of Turku Master’s Degree Programme in Public Mental Health is a 120 ECTS, two-year Master of Science programme in the Faculty of Medicine, Research Centre for Child Psychiatry. The controlling Peppi object is MDPPMH2427 / programme 98815. The exact thesis is PMH0020 Master’s Thesis in Public Mental Health, 40 ECTS, an Advanced Studies course in English graded 0–5. It is based on practical research and analysis of Public Mental Health data. The course estimates about eight months for hands-on research plus writing and defines a 1,080-hour workload. Most importantly, the thesis plan must be accepted by the PMH Master Programme Board before hands-on work begins.

1. Read the 120 ECTS degree and Peppi ranges correctly

The public programme page defines the degree as exactly 120 ECTS: 40 ECTS Major Studies and Research Methods, 40 ECTS thesis, 20 ECTS minor studies and 20 ECTS electives. Peppi represents the major-plus-thesis block as PMH0100, 80 ECTS, but its root can display 115–125 ECTS because minor/elective choices are represented as ranges. Do not describe the degree as 115 or 125 ECTS. Use your current personal study plan to resolve the exact minor/elective combination that brings the degree to 120.

2. PMH0020 is the exact thesis course

PMH0020 is 40 ECTS, Advanced Studies, English, and graded 0–5. The course expects a research project and an article-style scientific report rather than a generic essay. The stated structure is Introduction, Aims of the study, Methods, Results and Discussion. The programme description separately requires scientific thinking, mastery of necessary methods, familiarity with the topic and scientific English.

3. The PMH Programme Board gate is mandatory before hands-on work

At the beginning of thesis work, the student creates a thesis plan with supervisors. After submission, the PMH Master Programme Board must accept the plan before the student starts hands-on work. This is a programme-specific gate. Do not substitute the TBMC board process from Human Neuroscience, the responsible-professor sequence from Biomedical Imaging or the DDD workflow. Resolve the question, data source, methods, permissions, ethics, analysis plan and practical feasibility before this gate.

4. Plan around the actual 1,080-hour workload

PMH0020 allocates 10 hours of lectures, 40 hours of assignments, 30 hours of thesis-seminar presentations, 200 hours of reading and 800 hours of thesis research/writing. That totals 1,080 hours. The course estimates approximately eight months for hands-on research plus writing. Treat that as an integrated research period in which literature, data work, analysis, seminars and writing overlap, not as eight months of uninterrupted data collection.

5. Thesis seminars are inside PMH0020

PMH0020 explicitly requires active participation in Master’s thesis seminars and includes 30 hours of seminar-presentation work in the thesis workload. Planning begins during first-year research seminars, while implementation and reporting occur mainly in the second year. This means seminar activity is part of the thesis process itself.

6. PMH0011 is not a separate 4 ECTS thesis seminar

PMH0011 Orientation to University Studies and Seminars, 4 ECTS is a separate Pass/Fail course in the non-thesis part of the 80 ECTS PMH0100 block. It covers orientation, study planning, support services, wellbeing and general research ethics. Its title contains “Seminars,” but the current evidence does not justify calling those four credits a separate thesis-seminar package. PMH0020 already contains the actual Master’s thesis seminars.

7. Choose a project that fits the programme’s research environment

The Research Centre for Child Psychiatry runs national and international projects on mental health and psychosocial wellbeing of children, adolescents and families. Students can choose topics from ongoing or starting projects, and PMH0020 says topics are selected with PIs from the Child Psychiatry Research Centre and INVEST groups. A strong thesis therefore starts from a feasible project question and a real data or research environment, not from a broad topic such as “improve youth mental health.”

8. Understand the supervision and examiner chain

The programme description states that a person with at least a higher university degree can supervise, while the formal examiner route is stricter: at least two examiners, both with PhDs; one is an expert outside the research group and the supervisor acts as the other examiner. The Research Centre Board appoints examiners under authorization from the Faculty of Medicine dean. The head of Clinical Department decides acceptance and the final grade based on examiner opinions.

9. Start every thesis with a design label

Public Mental Health spans observational epidemiology, clinical/intervention statistics, qualitative research, implementation research, digital interventions, systematic reviews and humanitarian research. Before analysis, label the design precisely: cross-sectional, cohort, case-control, randomized trial, non-randomized intervention, qualitative interview study, mixed-methods study, systematic review/meta-analysis, register study or another defined design. The design determines what claims are defensible.

10. Association is not causation in epidemiology

PMH0004 teaches measures of occurrence and association, bias, confounding, study designs and causality. A statistical association between exposure and mental-health outcome does not by itself prove that changing the exposure will change the outcome. State temporal order, plausible confounders, measurement limits and selection mechanisms. Use causal language only when the design and assumptions support it.

11. Cross-sectional and longitudinal evidence answer different questions

Cross-sectional data can estimate patterns at one point or period but generally cannot establish whether exposure preceded outcome. Cohort designs can establish temporal ordering more directly but still face confounding and attrition. Case-control studies can be efficient for rare outcomes but depend heavily on case definition and control selection. Do not treat all “epidemiological data” as one evidence level.

12. Prevalence, service use and diagnosis are different constructs

Population prevalence, screening-score prevalence, diagnosed cases and service-use rates are not interchangeable. Access barriers, help-seeking, referral systems and administrative coding affect observed service data. If the thesis uses register or service-system data, define the outcome operationally and avoid presenting utilization as if it were direct population prevalence.

13. Mental-health measures need validity at the actual population level

A questionnaire or scale measures a defined construct under particular validation conditions. It is not automatically a clinical diagnosis. Report instrument version, language, scoring rule, cut-offs if used and relevant validation evidence. When studying children, adolescents, families or culturally different populations, consider whether the instrument performs similarly in the group being analysed.

14. Clinical-trial statistics require a pre-specified question

PMH0005 covers hypothesis formulation, randomization, outcomes, effect size, sample-size calculation, ANCOVA, pre-post analysis and CONSORT. For intervention research, define the primary question, groups, time points, primary outcome and analysis before looking for the most favourable result. Secondary and exploratory outcomes can still be useful, but label them honestly.

15. Pre-post improvement is not automatically intervention efficacy

Symptoms or wellbeing can change because of regression to the mean, natural recovery, secular trends, measurement effects or concurrent care. A pre-post change in one group therefore does not automatically prove treatment effect. A defensible intervention claim depends on the actual comparator, randomization or allocation process, baseline balance, attrition and analysis.

16. Report effect size and uncertainty, not only significance

A small p-value does not tell the reader whether a mental-health effect is large enough to matter. Report effect estimates and uncertainty where appropriate. Explain the scale and direction. If the sample is small, a wide confidence interval may show that several clinically or policy-relevant possibilities remain compatible with the data even when the point estimate looks promising.

17. Attrition can change who the intervention result represents

Digital and psychosocial interventions often lose participants over time. Report recruitment, allocation, follow-up and analysis denominators. If dropout differs by group, baseline severity, age or engagement, discuss how that could bias results. “Completed the intervention” is not the same population as “was offered the intervention.”

18. Feasibility, acceptability and effectiveness are separate outcomes

PMH0007 explicitly distinguishes intervention development, feasibility, implementation and impact. PMH0008 separately asks students to assess use, feasibility, effectiveness, efficiency and usefulness of digital interventions. A programme can be acceptable but ineffective, effective in a trial but difficult to implement, or feasible yet too small to support efficacy conclusions. Name the outcome category precisely.

19. Implementation success is not the same as clinical benefit

Reach, adoption, fidelity, acceptability, feasibility, cost and sustainability can explain whether an intervention enters routine service. These implementation outcomes do not themselves show that symptoms or functioning improved. Conversely, a clinically effective intervention can fail in practice because services cannot adopt or sustain it. Keep implementation evidence and participant outcome evidence in separate result layers.

20. Digital engagement metrics need a mental-health endpoint

Logins, completed modules, time in app or response rate can describe engagement. They do not establish mental-health benefit. If a digital intervention thesis makes an effectiveness claim, connect the intervention to a validated outcome and an appropriate comparison or design. Also discuss digital access, language, disability, device availability and service context when they can affect reach.

21. Intervention version and delivery context are part of the method

For digital, school, family or service interventions, record the intervention version, core components, delivery channel, dose/exposure, facilitator or automation, implementation setting and relevant co-interventions. A result from an intensively supported research version may not transfer to routine care unchanged. If the intervention changed during the study, document the version boundary.

22. Qualitative research needs transparent sampling and analysis

PMH0010 covers interviews, focus groups, observation, coding, thematic analysis and interpretation. Define who was sampled and why, how recruitment occurred, what interview or observation guide was used and how data were coded. Qualitative credibility comes from transparent fit between question, sampling, data collection and interpretation, not from pretending the sample is statistically representative.

23. Themes are interpretations, not simple vote counts

A theme that appears in many interviews is not automatically more important than a less frequent but conceptually significant theme. Explain how codes became categories or themes, how contradictory cases were handled and whether coding was reviewed by more than one researcher. Preserve an audit trail from source material to final interpretation.

24. Quotations can identify people indirectly

Mental-health interviews often contain family relationships, diagnoses, service histories, schools, municipalities or rare events that can identify someone even if names are removed. Select quotations with privacy in mind, remove unnecessary contextual identifiers and do not distort meaning while anonymising. Public thesis text and controlled raw transcripts are different disclosure surfaces.

25. Researcher reflexivity belongs in qualitative interpretation

The interviewer’s role, professional background, assumptions and relationship with participants can shape what is said and how it is interpreted. A concise reflexivity statement helps readers understand that relationship. It does not weaken the research; it makes the analytic process more transparent.

26. Systematic review work needs a reproducible search trail

PMH0009 teaches protocols, systematic searching, appraisal and meta-analysis. Preserve databases, search dates, search strings, eligibility criteria, screening decisions and extraction rules. If the review question changes after screening begins, record the change. A review is not systematic merely because it cites many papers.

27. Do not pool clinically different studies just because software can

Before meta-analysis, assess whether populations, interventions/exposures, outcomes and designs are sufficiently comparable. Statistical heterogeneity is only one part of the problem. A precise pooled estimate can still be misleading when studies answer materially different questions.

28. Meta-analysis does not convert association into causality

Pooling observational associations can improve precision but does not remove confounding or bias shared across studies. Similarly, a meta-analysis of small heterogeneous trials can remain uncertain. Interpret the pooled evidence at the level supported by the contributing designs.

29. Humanitarian settings require context-sensitive inference

PMH0002 addresses mental-health assessment, intervention and service planning in disasters and humanitarian crises. Cultural context, displacement, safety, language, disrupted services and acute needs can change both measurement and intervention feasibility. An instrument validated in one stable population should not automatically be assumed equivalent in a crisis-affected population.

30. Vulnerability changes consent and burden planning

Research with children, adolescents, distressed families or disaster-affected populations can involve heightened vulnerability. Minimise burden, use age-appropriate information, resolve consent/assent arrangements and consider whether questions may cause distress. Scientific value does not remove the duty to design a study proportionately.

31. Human-sciences ethical review is conditional, not automatic

University of Turku lists specific circumstances requiring human-sciences ethical review, including deviations from informed consent, interventions in physical integrity, certain research involving children under 15, exceptionally strong stimuli, mental harm beyond daily-life limits and safety threats. Review must occur before data collection when required. A master’s thesis is not automatically exempt and is not automatically required to undergo review solely because the topic is mental health.

32. Medical research and human-sciences review are different routes

Some health research falls under medical-research regulation; other behavioural, social or non-invasive health research may follow human-sciences review. Public Mental Health sits across those boundaries. Determine the legal and ethical category from the actual design, intervention, participants and data, not from the Faculty name or the word “clinical” in a dataset.

33. Register data can still need permissions and privacy controls

Register or administrative data may contain sensitive personal information. Ethical review requirements, data-holder permissions and data-protection duties are separate questions. Define the lawful access route, data controller/processor roles where applicable, permitted variables, secure analysis environment and whether linkage creates additional re-identification risk.

34. Pseudonymised mental-health data remain sensitive

Removing direct names does not automatically anonymise a dataset. Rare diagnoses, dates, service units, location, family structure or combinations of demographic variables can permit re-identification. If a key exists, pseudonymised data remain linkable. Limit access and do not copy protected data into uncontrolled tools.

35. AI use does not transfer responsibility

AI may support language, code or exploratory work where current rules and the project permit it, but the student remains responsible for accuracy, confidentiality and authorship. Do not upload sensitive participant, service, clinical or unpublished project data to uncontrolled AI services. Verify generated references against original sources and rerun code on controlled data before trusting output.

36. Preserve a source-population-to-figure provenance chain

For each headline result, trace the source population or dataset through eligibility, exclusions, missingness, variable derivation, analysis dataset, model and final figure/table. For qualitative work, trace interview guide, transcript/coding state and theme development. For systematic review, trace search and screening. For interventions, trace protocol and intervention version. This makes the thesis auditable across very different methods.

37. Separate exploratory from confirmatory analysis

Public-mental-health datasets invite many subgroup, endpoint and model choices. If a result emerged after inspecting the data, label it exploratory. Preserve the original plan and document changes. When possible, use sensitivity analyses to show whether the conclusion depends on one recoding, exclusion rule or model specification.

38. Null and negative results remain scientific results

An intervention can fail, a disparity can disappear after adjustment, a meta-analysis can remain inconclusive and qualitative data can challenge the expected theory. Do not suppress inconvenient evidence because it conflicts with a service, policy or partner preference. Explain whether a null result reflects imprecision, weak intervention exposure, measurement limits or genuinely little evidence of effect.

39. Turnitin and UTUGradu are separate final controls

University guidance makes Turnitin part of degree-thesis originality checking. UTUGradu manages higher-degree thesis submission, examination, approval, publication and archiving. Turnitin does not validate causal inference, intervention design or qualitative analysis. Scientific validity, research integrity and originality are separate obligations.

40. Final Public Mental Health checklist

Confirm MDPPMH2427 / programme 98815, the 120 ECTS degree, exact PMH0020 40 ECTS, 0–5 grading, the 1,080-hour workload, approximately eight-month research-and-writing estimate and the PMH Master Programme Board approval before hands-on work. Confirm that PMH0011 is not being double-counted as a separate thesis seminar. Then verify design label, outcome definitions, confounding/randomization logic, intervention and implementation endpoints, qualitative audit trail, review/meta-analysis provenance, ethics, permissions, personal-data controls, AI use, Turnitin, UTUGradu and public-thesis confidentiality.

Final evidence and reproducibility audit

Before candidate freeze, create a one-page evidence map for the thesis. For every headline conclusion, record the design, analytic sample, exposure or intervention, outcome, comparison, estimate and uncertainty, then state the strongest defensible verb. “Associated with,” “predicted,” “was acceptable,” “was feasible,” “reduced symptoms in the randomized comparison,” and “was implemented with high fidelity” describe different forms of evidence. This audit prevents an observational association from becoming an intervention claim or an implementation outcome from becoming a clinical-effectiveness claim during final editing.

Create a participant or record flow that matches the analysis file. Start from the source population, invited sample or records available; then document eligibility, consent where relevant, exclusions, duplicate removal, missing outcome data and the final denominator for each major analysis. If denominators change between tables, explain why. For repeated observations or clustered service data, record the level at which observations are independent and make sure the statistical model reflects that structure.

For intervention work, preserve the protocol and intervention version used for each participant or implementation site. Report departures from intended delivery, co-interventions, staff changes or technical outages when they can affect interpretation. If the project measures both implementation and mental-health outcomes, keep the corresponding datasets and analytic decisions separate enough that a reader can see whether poor outcome reflects an ineffective intervention, low exposure, implementation failure or insufficient statistical precision.

For qualitative work, retain a controlled audit trail linking the interview or focus-group guide to transcripts, coding, codebook revisions and final themes. Record when the coding framework changed and why. If translations were used, describe how meaning was preserved. Quotations should be checked against the transcript and then checked again for indirect identifiers before entering the public thesis. A compelling quotation is illustrative evidence, not a substitute for explaining the analytic process.

For systematic reviews and meta-analyses, rerun the final search close enough to manuscript freeze that the search date is honest and reproducible. Verify that every included study meets the stated criteria, each extracted effect uses the intended definition and the forest plot matches the extraction sheet. If risk-of-bias assessment changes, check whether the conclusion changes. A sensitivity analysis excluding high-risk studies can be more informative than a larger pooled sample.

Finally, rerun at least one headline result from the controlled source data through the final script or analysis workflow. Check units, labels, sample sizes and figure captions against the analysis output. Make sure the manuscript does not report an older exploratory value after the final dataset or model changed. This small reproducibility spot-check is particularly valuable in public mental health because projects often combine multiple data sources, repeated analyses, policy-facing outputs and late-stage formatting changes.