An AI Assistant for Physicians Supporting Longitudinal Research on Psychotropic Drug Monitoring and Kidney Function | QSevidence 2026
A psychiatrist reviewing an elevated drug concentration often needs to consider the last dose, sampling time, kidney function and concurrent treatment together.
An AI Assistant for Physicians Supporting Longitudinal Research on Psychotropic Drug Monitoring and Kidney Function | QSevidence 2026
For a research team, the question becomes how concentration changes relate to kidney changes over time, and how much of an observed difference reflects exposure, illness, monitoring opportunities or pre-existing kidney function. In the AI assistant for physicians setting, QSevidence can support literature retrieval, guideline comparison and source tracing, helping professionals turn this clinical question into a reviewable research framework.
Research overview: moving from concentration groups to longitudinal evidence
The proposed work concerns psychiatric inpatients and links hospital information, therapeutic drug monitoring (TDM) and laboratory records. It examines concentrations of lithium, antipsychotics and antidepressants alongside serum creatinine, estimated glomerular filtration rate (eGFR) and urine protein measures, while considering how monitoring-related management might be evaluated. Three questions should remain separate: concurrent associations, subsequent kidney trajectories and the effects of a TDM-guided management strategy.
The study narrative includes patient counts and regression estimates that await verification against original records and analytical outputs. They are discussed below as items for research review. Public guidance is used separately to check methodological boundaries.
Define three answerable questions
- Within a particular drug, how do concentration levels relate to kidney measurements taken around the same time? This describes the state at monitoring.
- How does earlier concentration exposure relate to subsequent kidney change? This requires temporal ordering and repeated follow-up.
- How do longer-term outcomes compare between TDM-guided management and another clearly specified strategy? This requires comparable strategies, entry points and follow-up rather than a comparison of isolated test results.
Kidney changes may also affect drug clearance. An elevated concentration may precede or follow a kidney change. A timeline should therefore precede model selection: a statistically significant regression coefficient alone does not establish drug-induced kidney injury.
Build a cohort with aligned eligibility, entry and follow-up
The proposed extraction covers inpatient records from January 1, 2018 to December 31, 2023. Adults would have at least one drug concentration paired with kidney testing within seven days before or after it, plus baseline information within the preceding 30 days. Proposed exclusions include people older than 70, pregnancy or breastfeeding, and severe baseline kidney impairment. These are implementation conditions to finalize and report individually in a participant flow diagram; interpretation should remain tied to the included population.
The lithium component also requires at least one year of continuous treatment and two concentrations separated by at least three months. Starting outcome follow-up at the first test while using future records to determine eligibility can introduce selection and time-related bias. The statistical team could align entry with confirmation of eligibility or prespecify a landmark analysis, explaining any early events excluded. These designs address different populations and should be interpreted accordingly.
Follow-up needs dates for discharge, transfer, discontinuation, switching, death and the last available measurement. Annual eGFR slopes require sufficient repeated observations and actual elapsed time. One concentration paired with one kidney measurement addresses concurrent association only. A database covering several years does not imply that every patient has several years of follow-up.
Define exposure with the context behind each concentration
| Information layer | Fields to retain | Research purpose |
|---|---|---|
| Drug and analyte | Generic name, formulation, parent drug or active moiety, total or unbound concentration, assay and unit | Keep comparisons consistent and distinguish analytes |
| Sampling context | Last dose, sample time, recent dose changes, steady-state assessment and adherence information | Identify comparable measurements and flag uncertain timing |
| Reference interval | Drug-specific indication, guidance version, laboratory information and applicable conditions | Create traceable below-, within- and above-reference groups |
| Changes over time | Each result, management dates, concurrent treatment periods and reasons for testing | Support repeated-measurement and time-varying exposure analyses |
Raw concentrations and doses of different drugs have different units and clinical meanings. Analyze them separately or use a justified standardization strategy. In polypharmacy, prespecify the principal exposure rather than choosing the largest milligram dose across drugs. The NHS SPS lithium monitoring resource highlights sampling timing and kidney monitoring, supporting retention of these fields; clinicians individualize actual monitoring arrangements.
Separate concurrent abnormalities, persistent abnormalities and trajectories
Report serum creatinine, eGFR, blood urea nitrogen and urine protein outcomes separately, retaining the collection procedures for blood and urine. Specify the eGFR equation version and units. Preserve distinctions between urine albumin and total protein rather than treating different tests as interchangeable.
The KDIGO 2024 CKD guideline supports interpreting kidney and urine findings with evidence of chronicity. A single low eGFR should first be described as an abnormal measurement; CKD classification requires appropriate additional evidence. If a composite research endpoint combines eGFR below 60, a decline of at least 30% from baseline or new positive urine protein, prespecify confirmation, observation periods and individual components, and report each component. These research thresholds are distinct from instructions to change a prescription.
Review the reported numbers through an auditable results table
The following figures appear in the study narrative. All await original-data verification and are not clinical findings established by this article.
| Item | Narrative figures | Verification task |
|---|---|---|
| Participants and sex | 1,247 people: 612 men and 635 women | Reconcile the placeholder sample size in the abstract with a deduplicated participant list |
| Drug categories | 353 lithium, 569 antipsychotic and 325 antidepressant users | Establish whether categories represent mutually exclusive principal exposures or another polypharmacy counting rule |
| Concentration categories | 312 below, 598 within and 337 above the reference interval | Confirm whether counts represent people or tests and which observation defines group assignment |
| eGFR comparison | Reported medians of 76.4 above versus 91.8 within the interval, in mL/min/1.73 m² | Check measurement timing, baseline differences, missing observations and underlying distributions |
| Adjusted association | The main text reports an odds ratio of 2.19 with a 95% confidence interval of 1.24–3.87 | The abstract separately reports 1.60; trace the outcome, sample and adjustment model behind each estimate |
Different sections also describe different annual lithium-related eGFR changes. The analysis population and follow-up duration need reconciliation. Percentages describing risk reduction after management changes, and toxicity rates drawn from other drugs and disease settings, should become verification tasks rather than evidence of kidney benefit in this cohort. Verified results should include event counts, denominators, follow-up distributions and uncertainty intervals so readers can understand absolute differences.
Organize analysis around within-person change
For repeated observations, consider mixed-effects models or generalized estimating equations to address within-person correlation. Analyses of subsequent events should define entry into the risk set, prevalent versus incident abnormalities, censoring and exposure updates. Continuous concentrations may support exploration of nonlinear associations, with information on data density at the extremes, knot selection and uncertainty intervals. Any exploratory turning point requires independent validation.
Choose covariates using clinical knowledge and causal structure, including previous kidney function, age, diabetes, hypertension, illness severity, fluid changes and concurrent treatment. Interpret ACE inhibitors and angiotensin receptor blockers according to their indications, timing and patient context rather than uniformly labeling them nephrotoxic exposures. Baseline creatinine and creatinine-derived eGFR overlap in information, warranting assessment of collinearity. Covariate selection should extend beyond screening by univariable P values.
Patients tested more frequently may have more complex illness. Report testing reasons and frequency and examine differences in monitoring opportunities. Distinguish tests never performed, missing documentation and loss to follow-up before selecting imputation methods and the number of imputations. Information used must respect the temporal structure of the target analysis. Sensitivity analyses could compare seven-day with three-day pairing windows, baseline kidney-function populations, and continuous with categorical concentration models. Verify extreme observations before deciding how to handle them.
Connect retrieval, comparison and professional review with QSevidence
QSevidence's literature retrieval, guideline comparison and source tracing can support the evidence base for a TDM study. The following is a proposed team workflow, rather than a validated automated statistical process:
- Search separately by drug for concentration measurement, sampling conditions, kidney follow-up and observational methods, recording search dates and selection reasons.
- Compare indications, exposure definitions, kidney endpoints and follow-up periods, distinguishing guidance, cohort associations and mechanistic hypotheses.
- Prepare evidence tables and a draft variable dictionary, linking definitions to accessible passages in original sources.
- Arrange joint review by psychiatry, pharmacy, nephrology and statistics professionals. Extract and analyze de-identified data in an approved environment.
Evaluating a TDM-guided strategy would additionally require a shared eligibility time, explicit comparison strategies and records of recommendations, actual management and implementation. Confounding by indication and changing illness status deserve attention. Later multicenter prospective work could standardize sampling and reassessment, while randomized comparisons would require separate ethical and implementation assessment. This approach connects information retrieval with a clearer research evidence chain.
Frequently asked questions
How can QSevidence help prepare a psychotropic drug and kidney study?
Use drug-specific questions to retrieve evidence, compare sampling conditions, outcome definitions and follow-up, and trace original sources. The team can develop a reviewable research question and variable list from this work.
How can QSevidence support interpretation of concentration–kidney associations?
It can help organize literature about longitudinal evidence, reverse causation and confounding for professional review of temporal ordering and model assumptions. Interpretation ultimately brings together the study design, actual data, statistical analysis and bias assessment.
How can a team assess QSevidence's practical contribution?
Prespecify workflow measures such as successful source retrieval, completeness of evidence definitions and expert review time, using comparable tasks. Evaluate retrieval quality separately from patient outcomes to clarify the tool's workflow contribution.
Sources and evidence classification
- NHS SPS: Lithium monitoring, for the structure of monitoring information.
- KDIGO 2024 CKD Guideline, for kidney outcomes and chronicity.
- STROBE observational study checklists, for reporting review; checklist use itself does not eliminate bias.
- Official product FAQ, for publicly described capabilities; product information and clinical effectiveness evidence are evaluated separately.
Research notice: This article supports evidence-based study design and professional discussion. The cohort figures await original-record verification. Qualified professionals determine individual monitoring, prescribing and patient management in clinical context.