Cumulative level test: Over time and at scale: course, prognosis and the delivery of care that holds

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 paper reports prevalence of a psychotic disorder as 0.4% in a region and incidence as 30 per 100,000 person-years. A reader computes an implied average duration and objects that the two figures are inconsistent. What does lesson 81 say about the relationship?Prevalence approximates incidence multiplied by average duration, so at 30 per 100,000 per year a 0.4% prevalence implies an average duration of roughly 13 years, which is plausible for a disorder with long-term disability and is a check on whether the two estimates were measured on comparable populations.Prevalence and incidence are independent quantities, so no inference follows from comparing them and the reader's objection is arithmetically confused.

Question 2

A fictional model flags psychiatric inpatients at high risk of post-discharge suicide. It has an area under the curve of 0.71 in external validation and its calibration slope is 1.0. A clinician says this means the model can tell her which of her patients will attempt suicide. What is the precise reading?An AUC of 0.71 with perfect calibration means about 71% of attempts will occur in flagged patients, so the flag identifies the patients who will go on to attempt.The model discriminates modestly and is well calibrated on average, which means predicted risks match observed rates across groups; but at a low base rate most flagged patients will not attempt, so it can support a conversation and a care plan, not an individual prediction.

Question 3

A fictional patient with recurrent depression has remitted on treatment for the second time in four years and asks whether she must stay on medication. Her clinician cites a trial in which continuing treatment produced a relative risk of relapse of 0.40 at one year with an absolute risk reduction of 25 percentage points. What should the discussion contain, on lesson 83's terms?The absolute risks with their time frame, the number needed to treat for one year, the harms and burdens of continuing, the reasoning behind her individual risk (previous episodes, residual symptoms), and the fact that stopping trials enrol stabilised, adherent patients, so her own course may differ.The 60% relative reduction should be quoted, since absolute risks are a simplification that understates the benefit of continuing treatment in a person with recurrent illness.

Question 4

A fictional guideline writer reads STAR*D and proposes to state that after four treatment steps, two thirds of people with depression achieve remission, and that the sequence demonstrates the value of each successive strategy. What must the writer confront?The cumulative figure is the correct summary of the trial and the only caution needed is that remission was measured with a self-report scale rather than an interview.The cumulative figure comes from an open-label sequential trial with substantial attrition at each step, decreasing per-step remission (about 37%, 31%, 14% and 13%), no placebo and no randomised comparison of strategies, so it describes what happened in that cohort rather than the causal value of sequencing.

Question 5

A fictional mental health charity campaigns on the claim that mental disorders now account for a rising share of global disease burden, citing the increase in disability-adjusted life years from 80.8 million in 1990 to 125.3 million in 2019. A critic replies that age-standardised rates barely moved, from about 1581 to 1566 per 100,000. Who is right, and how should both figures be used?Both are correct and answer different questions: the absolute burden rose mainly because populations grew and aged, while age-standardised rates were stable; the rising share of total burden reflects both population change and the relative decline of other causes, so the claim needs its denominator spelled out.The critic is right, so the campaign's claim is false: if age-standardised rates are stable then the number of affected people cannot meaningfully have increased.

Question 6

A fictional trial of a psychological treatment for panic reports large effects, but eligibility required no comorbidity, delivery was by doctoral-level therapists with session recordings and weekly supervision, sessions were individual and 16 in number, and the primary outcome was a symptom scale analysed in completers. A service lead wants to know whether to adopt it. What is the right conclusion?The trial shows the treatment works and the service lead's only task is to purchase training, since therapist effects and delivery conditions are second-order concerns once efficacy is established.The trial is explanatory on eligibility, organisation, flexibility of delivery, follow-up and analysis, so it shows the treatment can work under favourable conditions; adoption depends on whether the service can reproduce the delivery conditions, which is an implementation question the trial does not answer.

Question 7

A fictional service introduces session-by-session symptom ratings with a rule that a score above a threshold triggers a supervision review. An audit finds clinicians complete the measure in 55% of eligible sessions, and that completion is lowest for the patients with the highest scores. What does lesson 87 imply about how to interpret the programme's results?The programme's outcome data are differentially missing exactly where the intervention matters, so any improvement estimate is biased by non-completion of the highest-risk group, and the fidelity of completion is itself an implementation outcome that has to be measured and addressed.The data are usable because 55% is a majority of sessions, and missingness at that level is handled by the routine statistical adjustments available in any outcome analysis.

Question 8

A fictional evaluation of a parenting programme delivered across 20 sites reports that sites with high fidelity had substantially better child outcomes than sites with low fidelity, and concludes that the programme works when delivered well. What is the strongest methodological objection, and does it invalidate the finding?The comparison is invalid because fidelity was measured by the programme's own developers, and self-reported adherence is worthless in implementation research.Fidelity is not randomly assigned: high-fidelity sites may differ in staffing, leadership, caseload and population, so the association is confounded; it does not invalidate the finding but it means the comparison supports a hypothesis about implementation rather than an estimate of the effect of fidelity.

Question 9

A fictional blended service pairs a therapist with an app that sends weekly symptom ratings and prompts between sessions. In a pilot, therapists report that they use the dashboard in about a third of sessions, and patients with the highest scores are the least likely to submit ratings. Using lessons 87 and 89, what is the main risk to the model and what would you measure?The model depends on the data reaching the clinician before the session and on the patients most in need supplying it; the risks are silent disengagement and feedback that arrives too late or not at all, so the service should measure submission by severity, dashboard use per session, and what changes when a score crosses threshold.The main risk is data security, since any digital system holding symptom ratings creates a confidentiality problem that outweighs questions about engagement.

Question 10

A fictional paper states that a collaborative care model 'has been shown to be effective, with a pooled standardised mean difference of 0.30 for depression outcomes'. A reader wants to know whether that number settles the question of whether to fund it locally. Using lessons 86 and 90, what else is needed?Nothing else is needed: a pooled effect from randomised trials is the highest form of evidence, and implementation questions are matters of management rather than evidence.The number establishes a modest average benefit in the trials that were done; funding requires the comparison conditions, the population treated, the components that produced the effect (care manager, supervision, registry, measurement-based adjustment), the workforce available locally, and what activity would be displaced.

Question 11

A fictional student writes: 'Prevalence of depression in the clinic was 45%, showing that depression is far more common than community surveys suggest (about 5%). This confirms that screening is the answer.' Identify the errors using lessons 81 and 82.A clinic prevalence is a selected prevalent pool, not a population estimate; and the inference that a high proportion of attendees implies under-detection ignores that attendees were referred, that duration inflates clinic counts, and that a screening programme's value depends on what follows a positive result.The only error is the comparison of a clinic figure with a community figure, which should have been made against another clinic population of similar composition.

Question 12

A fictional director is asked to decide between two uses of the same budget: extending maintenance treatment for 200 people with recurrent psychosis, or funding a digital self-management app for 5,000 people with mild to moderate anxiety. Both have evidence. What does the sequence of lessons 81 to 90 suggest the decision requires?The comparison is straightforward because both interventions have randomised evidence: the option with the larger pooled effect size relative to control should be funded.A common currency is needed: expected events prevented per pound in each option, which requires absolute effects with their time frames, the treated proportion actually reached (adherence to maintenance, engagement with an app), harms and burdens, and the population denominator, rather than comparing one trial's effect size with another's.

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.