Cumulative level test: From lived experience to service systems: how psychiatric knowledge is made, measured and contested

Check understanding, separately from reading

12 questions · 80% to pass · no timer. This longer test revisits both units in the level. These authored questions assess recognition and application of the taught distinctions, not professional qualification. You may review the lessons and retry. Repeat attempts reuse the question bank; a remembered answer is not proof of transfer to a new situation.

Saving a lesson records reading only. Previous lesson completions have not been converted into passing scores. You can continue reading without a pass; the assessment remains unpassed.

Question 1

A fictional research team reports that a five-item experience scale has internal consistency of 0.91, test-retest reliability of 0.85, and a factor structure matching its published form. Their paper proposes the scale as a clinical screening tool for a specific perceptual anomaly. Using this level's treatment of description, instruments and measurement, what has the team established and what has it not?It has established consistency and structure: the items hang together, the scores repeat when nothing has changed, and the scale behaves like a one-dimensional measure, but not that the scale codes the anomaly as described, not that its scores behave equivalently across the groups it would be used with, and not what a score or a change in it means.It has established that the scale measures the anomaly, since a high internal consistency combined with a stable factor structure and reliable retest scores is the standard evidence that a scale is unidimensional and therefore specific to the construct it was designed for.

Question 2

A fictional early-intervention service proposes to use a phenomenological instrument to decide who receives a preventive intervention, citing a follow-up study in a high-risk sample in which the instrument's total score predicted shorter time to onset in a Cox model controlling for other predictors, and a diagnostic study in which scores aggregated in one spectrum. Write the reasoning that should decide whether the service can do this.Neither study design can be used because a Cox model estimated in a high-risk sample and a cross-sectional comparison of diagnostic groups are observational rather than experimental, and only randomised evidence can support a decision about who should receive an intervention.Neither study supplies what a selection rule needs: a threshold with measured sensitivity and specificity, the outcome base rate in a population like the service's, and the proportions above and below the cut-off who go on to develop the outcome, plus evidence that the instrument adds information beyond standard assessment. The prediction model was estimated in a small single cohort, and aggregation by group is not specificity for an individual.

Question 3

A fictional guidelines committee is choosing between a long-established diagnostic category and a proposed replacement. Evidence presented for the replacement: better agreement between raters (kappa 0.62 versus 0.45 under the study's own design), narrower criteria, and a supporting narrative from clinical experts. Evidence against: no follow-up or family study has compared the two. What should the committee's reasoning be?Agreement is a reliability result and does not adjudicate between the two categories on validity; the missing element is a head-to-head comparison under validators agreed in advance, course, family aggregation, treatment response, showing that the replacement performs better, and reliance on expert narrative reintroduces exactly the top-down expert classification that the validation programme was designed to replace.The replacement should be adopted because a kappa of 0.62 crosses the conventional threshold for substantial agreement while 0.45 sits only in the moderate range, and reliability is the primary evidence for a psychiatric diagnosis.

Question 4

A fictional trainee drafts a paragraph: 'Three meta-analyses report that this intervention works, with pooled effects of 0.4 to 0.6. Because synthesis is at the top of the evidence hierarchy, the intervention is established.' Rewrite the appraisal in this level's terms, naming the checks that determine how much the pooled figure can bear.The paragraph is sound in its structure; the remaining checks concern only the internal quality of the individual trials, such as blinding and allocation concealment, because synthesis removes the problem of how the included studies were selected.Pooled effects of 0.4 to 0.6 across three syntheses can be an arithmetic summary of a selected sample: the checks are whether unpublished and registered trials were sought, whether the trials were prospectively registered and their outcomes pre-specified, whether the number of studies permitted any assessment of small-study effects, whether the certainty of the estimate was rated with named reasons for rating down, and whether the three syntheses draw on overlapping studies so that 'three' may describe one literature rather than three independent bodies of evidence.

Question 5

A fictional guideline recommends an intervention on the basis of a 20% relative reduction in a serious event, and a service with a caseload spanning a wide range of baseline risks asks how to describe the benefit to two different people: one with an estimated two-year event risk of 50%, the other 5%. Compute the absolute effects and state what follows for how the recommendation should be presented.At a 50% baseline risk the absolute reduction is 10 percentage points, an NNT of about 10; at 5% it is 1 percentage point, an NNT of about 100. The same trial supports a materially different statement for each person, so the recommendation should be presented with absolute benefits at the relevant baseline risk rather than as a single relative reduction.At both risk levels the reduction is 20%, so the honest presentation is the relative figure, which is the trial's actual finding and therefore the most faithful thing to communicate to either person.

Question 6

A fictional formulary paper concludes: 'Certainty of evidence is low for this outcome, so the intervention does not work for it.' Which reformulation is correct, and what should the certainty rating have been doing?Low certainty is equivalent to no recommendation at all, so a formulary committee should defer the decision until certainty rises to moderate or high rather than reasoning about a low-certainty estimate.Low certainty means the estimate could easily be different; it may be compatible with a worthwhile benefit, with no effect, or with harm, and the rating should record which named factors were rated down (risk of bias, inconsistency, indirectness, imprecision, publication bias) so that a decision can weigh the plausible range rather than a verdict.

Question 7

A fictional service decides to adopt a therapy whose trials reported an effect size of 0.7 against waiting list. Six months after adoption, routine outcomes show no improvement relative to the service's previous year. List the explanations this level's evidence would put in order, and say what each implies for what to measure next.Start with the trial's comparator and delivery: the 0.7 estimate was against a waiting list, so part of it was non-specific improvement that routine care may already provide; then fidelity and staffing, since about 8.6% of outcome variance sits with the individual clinician and the trial's therapists were likely trained and supervised differently; then measurement, since routine outcomes may use a different instrument or a change threshold below the noise floor; and only then entertain that the treatment effect is absent.Conclude that the trial result was fraudulent or that the therapy does not work, since routine service data are the strongest available test of an intervention once it is in use.

Question 8

A fictional region has reduced inpatient beds by 15% and increased community team capacity. Two years later, admissions are down 18%, out-of-area placements are up 40%, and the number of people in contact with services is unchanged. Write the systems reading of those three numbers and the one measurement that would most clarify them.The pattern shows the community reform succeeded, since admissions are the outcome the reform targeted and they fell by more than the beds removed, while out-of-area placements are an administrative matter relating to bed availability elsewhere in the region.The pattern is consistent with redistribution and threshold shifting rather than reduced need: fewer local admissions and more placements outside the area suggests activity moved rather than disappeared, while unchanged contact numbers suggests the community capacity has not increased overall reach; the most clarifying measurement is the total number of bed-days used across all providers including out-of-area, plus the local admission criteria in each period.

Question 9

You are asked to appraise: 'A randomised trial of 300 patients found that a structured intervention reduced symptom scores by 4 points more than treatment as usual; the developers, who ran the trial, report that the difference exceeded the scale's minimal important difference; the intervention is recommended for routine use.' Identify the four decisive checks and the one that most changes the conclusion.Whether the minimal important difference estimate belongs to this instrument administered in this way at the relevant baseline severity; whether the trial was prospectively registered with the reported outcome pre-specified, given the developer's interest; whether treatment as usual contained the non-specific elements the intervention also supplies; and whether the delivery conditions, staff training, supervision, caseload, could be reproduced in routine services. The check that most changes the conclusion is the minimal important difference, because the entire claimed benefit rests on that threshold.Whether the trial's sample of 300 was large enough to detect a 4-point difference, since a p value below 0.05 demonstrates that the difference is real and the remaining questions are questions of interpretation rather than of whether the effect exists.

Question 10

Two fictional studies of the same diagnostic interview report agreement of 0.78 and 0.34. The second study used the same interview but had the second assessment conducted by a fresh interviewer without access to the first set of notes. A guideline developer asks which figure to use. What reasoning does this level supply, and what else about the interview would you want to know?The higher figure should be used, because the second study's method was less standard and fresh interviewers without access to previous notes cannot be expected to reach the same diagnosis reliably.Both figures are valid estimates of different things: agreement depends on the method, recorded review versus a fresh independent assessment, and on the heterogeneity of the sample, so the guideline should report both with their methods rather than choosing one; what is still needed is the distribution of the condition in each sample, since agreement is harder to achieve when prevalence is low, and whether the interview's categories were applied by clinicians trained to a common standard.

Question 11

A fictional clinician reads two reports of the same person's experience. The first records: 'Patient reports that thoughts are inserted; no insight.' The second records: 'Says thoughts arrive already spoken, in a voice that is not his, though he says they come from his own head; when asked whether that makes the thoughts his, says he does not know what would make a thought his.' Using this level's material, assess the two records and what each can support.The second preserves the person's words, separates the description from any judgement, and adds a distinction the first has hidden, the experience is estranged in its mode while its spatial origin remains his own, and his uncertainty concerns ownership rather than the fact of the experience, so it supports clinical reasoning about the form of the experience while the first supports only a conclusion, and neither supports a diagnosis.The first is the better clinical record because it is concise and uses standard terminology, which allows another clinician to reproduce the same assessment; the second is idiosyncratic and risks being misinterpreted by a reader who does not know the person.

Question 12

Assemble the reasoning for this decision: a fictional service with 4,000 people on its caseload wants to commission a new intensive intervention based on one positive randomised trial conducted in a different country, and the trial's own service had one trained clinician per 40 patients while this service has one per 120. Write the four-part assessment and the single cheapest measurement that would most reduce the uncertainty.The cheapest and most decisive step is to survey the caseload about their willingness to try the new intervention, since adoption is only worthwhile if the people concerned are willing to participate and satisfaction is the outcome most relevant to them.One: the evidence is a single trial, so provenance and pre-specification should be checked before the effect size is used at all. Two: the comparator and outcome definitions must match the local alternative, or the estimate will not transfer. Three: staffing is part of the intervention, one clinician per 40 versus one per 120 is a different treatment, so fidelity, caseload and supervision will determine whether the effect is even available here. Four: capacity and displacement decide reach: with limited capacity the intervention can serve a fraction of the caseload and will displace whatever currently fills those clinician hours. The cheapest uncertainty-reducing measurement is the service's own current outcome and contact data by clinician, which establishes the local baseline and shows how much clinician-level variance exists, allowing the trial effect to be read against local delivery conditions.

Continuing does not mark this assessment passed.

Progress stays in this browser, on this device, not an account or cloud backup. Clearing browser data removes it. Visits never count as study. Previous untimed legacy completions are not invented as study days.