Complete Academic Research Paper System Childhood Food Insecurity → Adult Cardiometabolic Health Given your design—a 20-year longitudinal cohort of 4,200 participants, existing ethics clearance, and a goal of submitting the first manuscript within 10 months—I would treat this as a causal-inference-oriented observational epidemiology paper, rather than a broad descriptive association study. The strongest positioning is: > Does childhood food insecurity predict cardiometabolic morbidity in adulthood, and to what extent are these associations explained or modified by subsequent socioeconomic, behavioral, and adiposity-related pathways? There is already evidence connecting childhood food insecurity with obesity and evidence from early-life nutritional adversity/famine studies linking childhood exposure with adult cardiometabolic conditions. A recent cohort study also found that early-childhood food insecurity was associated with worse cardiovascular health in young adulthood, primarily through BMI. Your opportunity is therefore not simply to demonstrate that food insecurity is associated with poor health. The publishable contribution should be the longitudinal characterization of timing, persistence, outcomes, dose-response, and/or pathways using your unusually long follow-up. --- 1. RESEARCH DESIGN AND QUESTION DEVELOPMENT 1.1 Recommended research problem The central problem should be framed around a limitation in existing epidemiologic evidence: Existing evidence establishes plausible links between childhood nutritional deprivation and later cardiometabolic risk, but important uncertainty remains regarding whether childhood food insecurity itself is associated with multiple adult cardiometabolic outcomes over the life course, independent of early socioeconomic disadvantage and other confounding factors. Your 20-year cohort provides several advantages: temporal ordering; repeated measurement; adult follow-up; ability to examine multiple outcomes; ability to distinguish persistent from transient exposure; ability to investigate potential mediating pathways; ability to assess differential effects across population subgroups. Avoid this framing > "Food insecurity causes obesity, diabetes, and hypertension." Your observational design cannot automatically establish that causal claim. Prefer > "We examined whether childhood food insecurity was associated with subsequent risk of obesity, type 2 diabetes, and hypertension over 20 years of follow-up." Then use causal language only if your analysis genuinely supports a causal interpretation. --- 1.2 Primary research question I recommend one primary question rather than treating three outcomes as three unrelated papers: > Among participants followed prospectively from childhood into adulthood, is childhood food insecurity associated with the subsequent development of obesity, type 2 diabetes, and hypertension? Secondary questions 1. Does the severity or frequency of childhood food insecurity show a dose-response relationship with adult cardiometabolic outcomes? 2. Do associations differ according to sex, race/ethnicity, childhood socioeconomic position, or other prespecified effect modifiers? 3. Are associations attenuated after adjustment for adult socioeconomic circumstances, health behaviors, or adult adiposity, and what does this suggest about potential pathways? 4. Does persistent or recurrent food insecurity confer greater risk than transient childhood food insecurity? 5. Are associations stronger for particular developmental periods of childhood/adolescence, if exposure timing permits? --- 1.3 Hypotheses Primary hypothesis Participants exposed to childhood food insecurity will have higher incidence of adult obesity, type 2 diabetes, and hypertension than participants who were food secure during childhood. Dose-response hypothesis Greater frequency/severity of childhood food insecurity will be associated with progressively greater adult cardiometabolic risk. Persistence hypothesis Persistent/recurrent childhood food insecurity will be associated with greater risk than a single episode. Effect-modification hypothesis Associations will differ across prespecified socioeconomic and demographic strata. Do not generate hypotheses for every variable in your dataset. That creates the appearance of post-hoc hypothesis generation. --- 1.4 Research gap identification Use a four-layer gap argument: Gap 1 — Outcome gap Much of the literature focuses on childhood obesity, whereas fewer studies simultaneously examine multiple adult cardiometabolic outcomes. A systematic review specifically identified the longitudinal literature on food insecurity and childhood obesity, illustrating how heavily the evidence base has concentrated on weight-related outcomes. Gap 2 — Life-course gap Many studies examine food insecurity contemporaneously with adult disease rather than establishing a clear childhood → adult temporal sequence. Gap 3 — Exposure-pattern gap A binary "food insecure vs secure" measure may obscure whether severity, duration, or persistence matters. Gap 4 — Mechanism/context gap The relationship may operate through several pathways: Childhood food insecurity → altered diet quality; stress/allostatic load; childhood adiposity; socioeconomic disadvantage; reduced physical activity; household instability; → adult obesity → diabetes/hypertension The key is to distinguish confounding variables from mediators. --- 1.5 Theoretical framework I recommend a life-course epidemiology framework, supplemented by a social determinants of health framework. Your conceptual model should be: Early-life socioeconomic conditions ↓ Childhood food insecurity ↓ Developmental/behavioral pathways nutrition stress growth/adiposity health behaviors ↓ Adult socioeconomic/behavioral environment ↓ Adult adiposity/metabolic dysfunction ↓ Obesity / T2D / hypertension This framework gives the paper substantially more theoretical coherence than simply listing socioeconomic variables in a regression model. Important distinction Create a DAG before finalizing the adjustment set. Do not automatically adjust for every variable associated with the exposure and outcome. For example, if childhood food insecurity contributes to childhood obesity and childhood obesity contributes to adult diabetes, childhood obesity may be a mediator, not a confounder. Similarly, adult BMI may be: a confounder for some outcomes; a mediator for others; or part of the outcome definition. The answer depends on the exact causal question. --- 1.6 Recommended design Your study is best characterized as: > Prospective longitudinal cohort study with secondary analyses potentially involving: survival analysis; repeated-measures analysis; mediation analysis; interaction/effect modification; competing-risk methods, if relevant. The primary analysis should be prespecified and relatively simple. Avoid turning Paper 1 into an omnibus statistical exercise. --- 1.7 Scope and delimitations Population 4,200 cohort participants originally enrolled during childhood/adolescence. Exposure Childhood food insecurity, operationalized according to the cohort's validated measurement instrument. Primary outcomes incident adult obesity; incident type 2 diabetes; incident hypertension. Primary estimand I would recommend: > Adjusted hazard/risk ratio comparing participants exposed versus unexposed to childhood food insecurity for incident adult cardiometabolic outcomes. The exact estimand depends on: outcome timing; repeated outcome measurement; censoring; exposure measurement; whether prevalence or incidence is available. --- 2. LITERATURE REVIEW FRAMEWORK 2.1 Search databases Use: Core PubMed/MEDLINE Embase Web of Science Scopus Supplementary PsycINFO CINAHL EconLit, if socioeconomic mechanisms become important --- 2.2 Search architecture Do not search: > "childhood food insecurity diabetes hypertension obesity" as one giant search. Build concept blocks. Block A — exposure "food insecurity" OR "food insufficiency" OR "household food insecurity" OR "food insecure" OR "food hardship" Block B — developmental period child* OR adolescen* OR youth OR "early life" OR "early childhood" OR "childhood" Block C — cardiometabolic outcomes obesity OR adiposity OR BMI OR "body mass index" OR diabetes OR "type 2 diabetes" OR hypertension OR "blood pressure" OR cardiometabolic OR metabolic Block D — longitudinal design longitudinal OR cohort OR prospective OR "life course" OR follow-up OR incidence Combine: > A AND B AND C AND D Then conduct outcome-specific searches. --- 2.3 Literature matrix Build a spreadsheet with: Field What to capture Citation Author/year Country Geographic context Cohort Study name N Sample size Age at exposure Childhood age Exposure definition Instrument Exposure timing Age/waves Outcome Obesity/T2D/HTN Follow-up Years Design Prospective/retrospective Statistical method Cox/log-binomial/etc. Covariates Adjustment set Effect estimate HR/RR/OR Main finding Direction/magnitude Limitations Authors' limitations Your assessment Methodological limitation Gap Contribution to your study --- 2.4 Literature synthesis Don't write: > Smith found X. Jones found Y. Brown found Z. That becomes an annotated bibliography. Instead organize by argument. Theme 1 — Childhood food insecurity as an exposure Define the construct and explain measurement differences. Theme 2 — Food insecurity and childhood adiposity Establish the strongest existing evidence. Theme 3 — Early-life deprivation and adult metabolic health Move from childhood outcomes to life-course consequences. Theme 4 — Diabetes and hypertension Identify evidence, limitations, and inconsistencies. Theme 5 — Mechanisms Discuss: diet quality; metabolic programming; stress; adiposity; socioeconomic pathways; behavioral pathways. Theme 6 — Methodological limitations Focus on: cross-sectional designs; retrospective exposure assessment; inconsistent food-insecurity measures; inadequate control of socioeconomic confounding; insufficient follow-up; attrition; failure to distinguish childhood from adult food insecurity. Theme 7 — Your contribution End with: > "The present study addresses these limitations by..." That sentence should effectively become the bridge into your research question. --- 2.5 Literature review target For a Q1 manuscript, I would not automatically write a 3,500-word dissertation-style literature review. Your requested 2,000–3,500-word structure is useful for the dissertation, but the journal manuscript needs compression. For the journal paper, aim approximately: Introduction + evidence synthesis: 700–1,000 words with the full literature review retained for your dissertation chapter. --- 3. METHODOLOGY DESIGN 3.1 Recommended analytical strategy Your analysis plan should be finalized before inspecting outcome-specific associations. I recommend the following hierarchy. Model 0 — Descriptive Compare: demographic characteristics; childhood socioeconomic characteristics; exposure prevalence; follow-up; outcome incidence. Model 1 — Minimally adjusted Childhood food insecurity + predetermined basic confounders. Model 2 — Socioeconomic adjustment Add childhood socioeconomic characteristics. Model 3 — Life-course adjustment Add prespecified adult socioeconomic/behavioral variables where justified. Model 4 — Mediation/pathway model Only if this is explicitly a secondary objective. Do not label Model 4 the "fully adjusted model" if it includes mediators. --- 3.2 Covariate strategy This is one of the most important parts of your paper. Create three categories: A. Confounders Variables that precede exposure and independently influence outcome. Potential examples: parental socioeconomic position; parental education; household income; childhood neighborhood; parental health characteristics; demographic characteristics. B. Mediators Potential pathway variables occurring after childhood food insecurity. Potential examples: childhood adiposity; adult diet; adult socioeconomic position; adult physical activity; adult BMI. C. Effect modifiers Variables where you genuinely hypothesize that the association differs. Potential examples: sex; race/ethnicity; childhood socioeconomic position; age at exposure. Build this using a DAG, not p-value-based variable selection. --- 3.3 Primary statistical models The optimal model depends on how your outcomes are recorded. If exact time to disease onset is available Use Cox proportional hazards regression. Report: > adjusted hazard ratios and 95% confidence intervals. Evaluate: proportional hazards assumption; influential observations; censoring; loss to follow-up. If outcomes are repeatedly measured Consider: discrete-time survival models; mixed-effects models; generalized estimating equations; marginal structural models where time-varying confounding warrants them. If only final adult status is available Use: modified Poisson regression with robust variance for risk ratios, where appropriate; logistic regression if odds ratios are substantively appropriate and unavoidable. Do not default to logistic regression simply because the outcome is binary. --- 3.4 Missing data This should be specified before analysis. Recommended: > Multiple imputation by chained equations under a missing-at-random assumption, provided the missingness structure and available auxiliary variables support the assumption. Report: proportion missing per variable; imputation model; number of imputations; variables included; diagnostics; comparison with complete-case analysis. Do not merely write: > "Missing data were excluded." That is likely to attract methodological criticism in a 20-year cohort. --- 3.5 Attrition A 20-year cohort inevitably raises attrition concerns. You should report: 1. baseline N; 2. eligible N; 3. N with exposure; 4. N with outcome data; 5. N included in primary analysis; 6. N lost to follow-up; 7. reasons where available. Compare participants retained versus lost to follow-up. If attrition is substantial, consider: inverse probability-of-censoring weighting; sensitivity analyses; selection-bias discussion. --- 3.6 Sensitivity analyses Pre-specify perhaps 3–5, rather than 15. Strong candidates: Sensitivity 1 Complete-case versus multiply imputed analysis. Sensitivity 2 Alternative definition of food insecurity. Sensitivity 3 Exclude participants with baseline cardiometabolic disease where appropriate. Sensitivity 4 Alternative outcome definition. Sensitivity 5 Inverse-probability weighting for attrition. Only include analyses that address plausible sources of bias. --- 3.7 Multiple outcomes You have three major outcomes. This creates a strategic issue. I would designate: Primary outcome: whichever outcome has the strongest a priori biological/public-health rationale and best measurement. Then: Secondary outcomes: remaining cardiometabolic outcomes. Alternatively, if your scientific question explicitly treats cardiometabolic morbidity as a multidimensional endpoint, you can prespecify all three as co-primary outcomes—but then address multiplicity carefully. Don't casually run three outcomes × six models × ten interactions and then highlight whichever produces p < .05. --- 3.8 Multiple testing You should explicitly distinguish: primary hypotheses; secondary hypotheses; exploratory analyses. For the primary three-outcome analysis, consider controlling the family-wise error rate or false discovery rate if appropriate to your inferential framework. At minimum, report effect sizes and confidence intervals, not merely statistical significance. --- 3.9 Effect modification Don't test every subgroup. Choose 2–3 biologically/public-health justified modifiers. For example: sex; childhood socioeconomic position; race/ethnicity, if theoretically justified and measured appropriately. Report interaction terms and stratum-specific estimates. Avoid: > "The association was significant in women but not men." That alone does not establish effect modification. You need evidence that the interaction/heterogeneity itself is supported. --- 4. COMPLETE PAPER STRUCTURE Given your target journals, I would modify your original word allocation. AJPM currently describes Research Articles—including observational studies—as having a structured abstract of ≤300 words, ≤3,000 words of text, and no more than four tables/figures. Preventive Medicine similarly specifies ≤3,000 words of main text, four tables/figures, 35 references, and a structured abstract ≤200 words for an Original Research Paper. So build the manuscript around the stricter 3,000-word architecture rather than producing an 8,000-word manuscript and cutting it later. --- Title Potential title: > Childhood Food Insecurity and Risk of Cardiometabolic Disease in Adulthood: A 20-Year Prospective Cohort Study Alternative: > Childhood Food Insecurity and Adult Obesity, Type 2 Diabetes, and Hypertension: A 20-Year Longitudinal Cohort Study The second is more explicit and probably stronger for discoverability. --- Abstract For AJPM: Objective 1–2 sentences. Methods Population, exposure, outcomes, follow-up, statistical model. Results Sample size, exposure prevalence, effect estimates, 95% CIs. Conclusions One primary interpretation + public-health implication. Never put unsupported causal language in the abstract. AJPM specifies a structured abstract of ≤300 words for Research Articles. --- Introduction — ~600–800 words Paragraph 1 Public-health significance of food insecurity. Paragraph 2 Childhood as a critical life-course period. Paragraph 3 Evidence connecting childhood food insecurity to obesity and cardiometabolic risk. Paragraph 4 Limitations of existing evidence. Paragraph 5 Your cohort's unique contribution. Final paragraph State: objective; primary hypothesis; secondary objectives. --- Methods — ~900–1,100 words Study design and population Describe: cohort; recruitment; baseline; follow-up; eligibility. Exposure Precisely define food insecurity. Outcomes Define: obesity; T2D; hypertension. Covariates Explain conceptual selection rather than simply listing variables. Statistical analysis Specify: primary model; adjustment sequence; missing data; attrition; sensitivity analyses; interactions; significance/CI approach. Ethics Briefly state approval and consent. Because your ethics clearance already exists, this section should be straightforward. --- Results — ~900–1,100 words Table 1 Baseline characteristics by childhood food-insecurity status. Table 2 Incidence/risk of obesity, T2D, hypertension. Table 3 Adjusted associations across model specifications. Figure 1 Participant flow diagram. Optional Figure 2 Adjusted effect estimates/forest plot. This gives you the four-table/figure ceiling relevant to AJPM and Preventive Medicine. --- Discussion — ~900–1,100 words Use five paragraphs/sections. 1. Principal findings Start immediately with the answer. 2. Comparison with previous research Explain: agreement; disagreement; why differences might exist. 3. Mechanisms Discuss plausible pathways without presenting speculative mechanisms as demonstrated mediation. 4. Strengths and limitations Strengths: longitudinal design; long follow-up; temporal ordering; validated measures; repeated observations; large cohort. Limitations: observational design; residual confounding; exposure measurement; attrition; generalizability; outcome ascertainment; potential misclassification. 5. Public-health implications Move from: finding → interpretation → intervention/policy implication rather than: finding → sweeping recommendation. --- Conclusion — ~100–150 words Answer: > What did we learn? Then: > Why does it matter? Don't introduce new findings. --- 5. ACADEMIC WRITING AND STYLE GUIDE 5.1 Academic voice Use: > "Childhood food insecurity was associated with..." rather than: > "Childhood food insecurity caused..." Use: > "These findings suggest..." rather than: > "These findings prove..." --- 5.2 Hedging hierarchy Strong evidence > demonstrates consistently observed robustly associated Moderate evidence > suggests is consistent with provides evidence that Observational uncertainty > may reflect could indicate is potentially explained by may be mediated through For your paper, "associated with" should be the default causal verb. --- 5.3 Evidence integration Weak: > Smith et al. found X. Jones et al. found Y. Brown et al. found Z. Strong: > Longitudinal evidence generally suggests that childhood food insecurity is associated with adverse weight-related outcomes, although findings vary according to exposure definition, developmental period, and socioeconomic adjustment. Then cite multiple studies. --- 5.4 Paragraph architecture Use: Claim → Evidence → Interpretation → Link Example structure: > Childhood food insecurity may represent an early-life exposure with consequences extending beyond childhood. Prior longitudinal studies have linked food insecurity with adverse weight trajectories... These findings suggest that nutritional deprivation may influence later cardiometabolic risk through both behavioral and physiological pathways. However, evidence linking childhood exposure specifically with adult diabetes and hypertension remains comparatively limited. That's the rhythm you want throughout. --- 5.5 Avoid these common errors Error 1 — Causal overclaiming "Food insecurity causes diabetes." Error 2 — P-value storytelling "There was a statistically significant association." Instead: > "Food-insecure participants had a 32% higher estimated risk..." with CI. Error 3 — Excessive covariate adjustment Throwing every available variable into the model. Error 4 — Post-hoc subgroup fishing Dozens of interactions with no theoretical justification. Error 5 — Reporting only adjusted estimates Show enough information for readers to understand how adjustment changed the association. Error 6 — Ignoring selection bias A 20-year longitudinal cohort will inevitably invite questions about attrition. Error 7 — Confusing mediation and confounding This is particularly important for BMI and adult socioeconomic variables. --- 6. TABLE AND FIGURE PLAN Table 1 Characteristics of participants according to childhood food-insecurity status Include: N; age; sex; race/ethnicity; childhood socioeconomic indicators; parental characteristics; relevant baseline health variables. Table 2 Incidence of adult cardiometabolic outcomes For each outcome: cases; person-years; incidence rate; crude association. Table 3 Adjusted associations between childhood food insecurity and adult cardiometabolic outcomes Columns: Model 1; Model 2; Model 3; HR/RR; 95% CI. Figure 1 Cohort flow diagram. Figure 2 Forest plot of primary and prespecified subgroup estimates, only if justified. --- 7. PUBLICATION STRATEGY Journal 1 — American Journal of Preventive Medicine This is arguably an excellent fit because AJPM explicitly publishes original observational studies and public-health investigations. Its current Research Article requirements include ≤3,000 words, structured abstract ≤300 words, and four tables/figures. Strategic pitch: > Early-life social exposure → preventable adult chronic disease → life-course prevention. That is a strong preventive-medicine narrative. --- Journal 2 — American Journal of Public Health AJPH is also conceptually attractive because the paper has an important social determinants/public-health equity dimension. For AJPH, emphasize: population health; inequity; childhood adversity; prevention; policy relevance. Don't frame it primarily as a metabolic physiology paper. --- 8. JOURNAL SELECTION — FIVE CRITERIA Score each journal 1–5 on: Criterion Weight Scope fit 30% Methodological fit 20% Novelty fit 20% Audience/policy relevance 15% Submission constraints 15% Your target journal should not simply be the one with the highest impact factor. Fit beats prestige when the objective is acceptance within 10 months. --- 9. REVIEWER-PROOFING Before submission, imagine three reviewers. Reviewer 1 — Epidemiologist Will ask: > What is the causal estimand? > Why were these covariates selected? > How did you handle missing data? > What about attrition? > Why Cox rather than another model? Reviewer 2 — Public-health scholar Will ask: > Why does this matter for population health? > What does it add beyond existing food-insecurity research? > What are the intervention implications? Reviewer 3 — Statistician Will ask: > Was the analysis prespecified? > How were multiple comparisons handled? > Were proportional-hazards assumptions tested? > Were interactions prespecified? > How sensitive are findings to missing data? Design the manuscript so these questions are answered before reviewers ask them. --- 10. STROBE REQUIREMENTS Because this is a longitudinal observational cohort, use the STROBE cohort checklist. STROBE specifically recommends reporting the study design, setting, eligibility, exposure/outcome definitions, bias considerations, sample-size rationale, missing data, loss to follow-up, sensitivity analyses, and adjusted/unadjusted estimates with precision. Treat STROBE as a submission-quality control system, not something you fill out at the end. [STROBE cohort checklist](https://www.strobe-statement.org/checklists/?utm_source=chatgpt.com) --- 11. YOUR 10-MONTH EXECUTION ROADMAP You have enough time—but only if you avoid spending months perfecting the literature review before analyzing the data. Month 1 — Lock the analysis [ ] Finalize causal question [ ] Construct DAG [ ] Define primary exposure [ ] Define primary outcomes [ ] Define confounders [ ] Define mediators [ ] Define effect modifiers [ ] Specify primary model [ ] Specify missing-data strategy [ ] Specify sensitivity analyses [ ] Freeze the analysis plan Deliverable: Statistical Analysis Plan v1.0. --- Month 2 — Data audit [ ] Validate variable definitions [ ] Check cohort eligibility [ ] Examine missingness [ ] Verify outcome ascertainment [ ] Examine follow-up [ ] Quantify attrition [ ] Validate exposure coding [ ] Create reproducible analytic dataset Deliverable: analysis-ready dataset + data dictionary. --- Month 3 — Primary analysis Run: 1. descriptive analysis; 2. primary model; 3. secondary outcomes; 4. sensitivity analyses; 5. prespecified interactions. Do not start inventing new hypotheses because an interesting coefficient appears. Deliverable: Results tables. --- Month 4 — Validation Perform: model diagnostics; sensitivity analysis; missing-data comparison; complete-case comparison; attrition analysis; alternative exposure definitions. Deliverable: locked results. --- Month 5 — Manuscript skeleton Write: Introduction; Methods; Results; Discussion outline. Do not polish prose yet. Deliverable: complete rough manuscript. --- Month 6 — Full manuscript Complete: abstract; tables; figures; references; supplementary material; STROBE checklist. Deliverable: Manuscript v1. --- Month 7 — Internal review Have: supervisor; epidemiology/statistics colleague; content expert review it. Ask them to identify: 1. causal overclaims; 2. methodological vulnerabilities; 3. unclear contribution; 4. unnecessary analyses. Deliverable: Manuscript v2. --- Month 8 — Journal targeting Submit first to your preferred journal. If choosing between AJPM and AJPH, make the decision based on the final findings, not before analysis. For example: Strongest finding Childhood food insecurity → adult cardiometabolic disease. AJPM framing: prevention/life-course risk. Strongest finding Large disparities by socioeconomic or demographic subgroup. AJPH framing: health equity/social determinants. --- Month 9 — Revision / resubmission If rejected without review: Do not rewrite the entire paper. Diagnose: scope mismatch; novelty; framing; word count; journal fit. Then transfer to Journal 2. If peer reviewed: Respond systematically to every reviewer comment. --- Month 10 — Final submission/revision Maintain: analysis code; version-controlled manuscript; response-to-reviewers document; supplementary files; STROBE checklist; cover letter. --- 12. RESPONSE TO REVIEWERS TEMPLATE Use this architecture: > Reviewer Comment: The reviewer argues that residual socioeconomic confounding may explain the observed association. Response: We thank the reviewer for highlighting this important consideration. We have clarified our confounding framework and additionally adjusted for [variables] in a prespecified sensitivity analysis. Change to manuscript: We revised the Methods section to clarify the covariate selection strategy (Page X, Lines X–X) and added the sensitivity analysis to Table Sx. Interpretation: The additional adjustment attenuated the association from X to Y but did not materially alter the direction of the association. Never respond defensively. The reviewer doesn't need to be "wrong" for you to improve the manuscript. --- 13. CONFERENCE PRESENTATION For a 10–12 minute presentation: Slide 1 Title + research question Slide 2 Why childhood food insecurity matters Slide 3 Knowledge gap Slide 4 Cohort/design Slide 5 Conceptual model/DAG Slide 6 Exposure/outcomes Slide 7 Primary results Slide 8 Secondary/sensitivity results Slide 9 Interpretation Slide 10 Public-health implications Slide 11 Strengths/limitations Slide 12 Take-home message Your final slide should contain one sentence, not six bullet points. --- 14. MOST IMPORTANT STRATEGIC RECOMMENDATION Because this is Paper 1 of three planned papers, do not use all the interesting variables now. Think of the dataset as a research program: Paper 1 Childhood food insecurity → adult cardiometabolic outcomes Primary longitudinal association. Paper 2 Mechanisms/pathways Potential mediation through: childhood/adolescent adiposity; diet; socioeconomic trajectories; stress/behavioral pathways. Paper 3 Heterogeneity/equity Who is most affected? socioeconomic position; sex; race/ethnicity; timing/persistence of exposure; resilience/protective factors. This prevents Paper 1 from becoming unfocused and preserves genuine novelty for Papers 2 and 3. --- 15. YOUR ANALYSIS PLAN — RECOMMENDED ONE-PAGE VERSION Before touching the final manuscript, I would have your supervisor sign off on this exact logic: Population: Eligible participants from the 20-year cohort with childhood food-insecurity information and appropriate adult follow-up. Exposure: Childhood food insecurity, defined using the cohort's validated measure. Primary outcomes: Incident adult obesity, type 2 diabetes, and hypertension. Primary estimand: Adjusted association between childhood food insecurity and each incident outcome. Primary model: Time-to-event model if onset dates are available; otherwise an appropriate risk model based on outcome ascertainment. Confounding strategy: DAG-informed, a priori covariate selection. Missing data: Multiple imputation, with complete-case sensitivity analysis. Attrition: Descriptive comparison + sensitivity analysis/IPCW if warranted. Effect modification: Limited, prespecified interactions. Sensitivity analyses: Alternative exposure/outcome definitions, missing-data strategy, and attrition. Inference: Effect estimates + 95% CIs; p-values secondary. Reporting: STROBE cohort guidance. That is the backbone of a defensible Q1 epidemiology paper. --- One final methodological caution The single most important thing to resolve before running the models is your causal structure. In particular, decide whether your scientific question is: A. Total effect > What is the association/effect of childhood food insecurity on adult cardiometabolic disease? or B. Direct effect > What is the association independent of subsequent adiposity and adult socioeconomic/behavioral factors? Those are different questions and require different adjustment sets. If your primary objective is the total life-course effect, indiscriminately adjusting for adult BMI, adult income, adult diet, and adult physical activity could remove part of the very pathway you are trying to estimate. For a PhD-level paper, I would therefore make the DAG + estimand + covariate strategy the first deliverable, before statistical modeling begins. That is where the scientific rigor of this project will be won or lost—not in whether you use Cox regression versus another regression model.