CRITICAL CARE MEDICINE THESIS TOPICS FOR MD/DNB

A critical care protocol is judged on internal consistency: the research question, objectives, methodology and statistical plan must all describe the same study. The primary objective should be a single measurable endpoint — one prevalence, one association, one comparison between two defined groups — with everything else demoted to secondary objectives.

Fix the time point. This is the defining requirement of this speciality and the one most often left out. Nearly every title on this list is labelled cross-sectional, but almost none of the variables are: lactate falls within hours of resuscitation, the Sequential Organ Failure Assessment score is recomputed daily, positive end-expiratory pressure is altered at the bedside, and a patient can move from hyponatraemia to normonatraemia overnight. A protocol that promises to correlate lactate with sepsis severity has not yet described a study. State the anchor explicitly: which value, measured at which moment relative to intensive care admission, and relative to which treatment. The usual conventions are the first value recorded within a defined window of admission, the worst value in the first twenty-four hours, or the value at the point of meeting the case definition — and it must be one of them, chosen and named, not left to whoever fills in the proforma. Where the exposure and the outcome are both measured, say whether they are measured at the same time point, because a genuinely cross-sectional design requires that they are.

Name the definitions with their versions. Sepsis and septic shock follow the current consensus definition, and the protocol must say which one, because a study using systemic inflammatory response syndrome criteria and a study using organ dysfunction criteria enrol substantially different patients — that difference is the whole point of the topics comparing them. Acute respiratory distress syndrome follows the Berlin definition, which requires the oxygenation ratio to be measured on a positive end-expiratory pressure of at least five, so the protocol must state that the ratio is recorded on the set pressure rather than pulled from any arbitrary blood gas. Acute kidney injury follows KDIGO staging, which needs a baseline creatinine that most emergency admissions do not have; state the surrogate you will use and apply it uniformly. Ventilator-associated pneumonia needs either the surveillance definition or a clinical one, not a mixture. Delirium follows a named validated instrument.

Say how the scores are computed and who computes them. The Acute Physiology and Chronic Health Evaluation II score uses the worst value of each variable in the first twenty-four hours; the organ failure score is computed per day. State the window, state the assessor, and state what happens when a component is missing — a bilirubin never sent, a blood gas not repeated. Scoring a missing variable as normal is the default in most software and it systematically underestimates severity, so if you do it, say so, and count how often it happened.

Address confounding by indication in the methodology, not only in the limitations. Several topics here relate a clinician-set variable to a patient outcome: positive end-expiratory pressure against oxygenation, mean arterial pressure against organ dysfunction, ventilator settings against underlying illness. These variables are chosen because of the patient's condition, so an observed association partly reflects the treatment tracking the severity rather than causing the finding. This does not invalidate the study, but the protocol must frame the objective as description or association rather than effect, and the analysis must adjust for the severity indication where possible.

A critical care synopsis is the condensed document of two to four pages — title, introduction, aim and objectives, brief methodology, sample size and references — submitted for registration of the dissertation topic. The critical care protocol is the expanded version of twelve to twenty pages carrying the full review of literature, detailed methodology including case definitions, time points, scoring rules and instruments, statistical plan, study timeline and annexures including the patient information sheet, consent form and data collection proforma.

Two annexures carry more weight in this speciality than elsewhere. The data collection proforma must be built around a fixed time grid rather than a list of variables, because the difference between a usable dataset and an unusable one is whether every patient's lactate was recorded at the same defined moment. Design the proforma with the time point printed against each field. And the consent form needs a surrogate version, since a substantial proportion of the patients you enrol will lack capacity at the moment of enrolment; a protocol submitted with only a patient-signature form invariably comes back from the committee.

In practice the synopsis is extracted from the protocol rather than written separately, which is faster and produces a more coherent document. Check your university's prescribed proforma before submission, since rejections on formatting grounds are common and entirely avoidable.

Match the formula to the design. Prevalence topics — delirium, acute kidney injury, thrombocytopenia, anaemia, electrolyte abnormalities, malnutrition — use a single-proportion calculation from a cited local estimate. Comparative topics between two defined groups need a two-proportion or two-mean calculation with both expected values referenced, not a single-proportion formula quietly reused. Correlation topics such as lactate against organ failure score use the expected correlation coefficient. Topics comparing two severity scores need a receiver operating characteristic calculation sized on the expected area under the curve and the anticipated event rate.

The score-comparison topics need a stated reference outcome. Comparing the quick organ failure score with systemic inflammatory response criteria, or with the full organ failure score, is only meaningful against something — usually in-hospital mortality, intensive care stay beyond a defined duration, or need for vasopressors. Name it in the objectives. Two consequences follow: the design acquires a short follow-up element even though the exposure is measured at a single point, which the protocol should acknowledge rather than conceal under the cross-sectional label; and the sample size is driven by the number of events, not the number of patients, so a unit with fifteen per cent mortality needs several hundred admissions to accumulate enough.

Name the tests. Intensive care data are rarely normally distributed — lactate, creatinine, duration of ventilation and length of stay are all right-skewed — so plan for medians with interquartile ranges and the Mann-Whitney U test as the default for continuous comparisons, with the normality test stated. Proportions use the chi-squared test with Fisher's exact test for sparse cells, which are frequent when individual organisms or reaction types are tabulated. Score comparisons use the area under the receiver operating characteristic curve with confidence intervals and a formal test for the difference between two curves, and a derived cut-off with its sensitivity and specificity where a threshold is proposed. Agreement between point-of-care ultrasonographic and clinical assessment of volume status is reported as kappa for categorical judgements and by Bland-Altman analysis for continuous measurements, never as a correlation coefficient.

Two arithmetic disciplines worth imposing on yourself. First, multivariable logistic regression needs roughly ten outcome events per variable entered; a model with twelve predictors and forty events is overfitted and its adjusted odds ratios are not interpretable, however impressive the table looks. Decide the predictor list in the protocol and justify its length against the expected event count. Second, state the unit of analysis. A patient may have several blood gases, several cultures and several transfusion episodes, and treating them as independent observations inflates the sample and narrows the confidence intervals falsely. If the unit is the episode rather than the patient, say how clustering is handled.

Finally, plan for missing data before you meet it. Intensive care records are dense but uneven, and a protocol that says nothing about missingness usually ends with the resident silently dropping incomplete cases, which is itself a selection process because the sickest and the briefest admissions are the ones with gaps. State the rule in advance.

Start with the admission register. Count bed strength, annual admissions and the case mix over the last twelve months, because that determines what accrues within your data collection window. A ten-bedded unit with roughly six hundred admissions a year supports a study of sepsis, acute kidney injury or electrolyte abnormality comfortably, since these affect a large share of admissions. The same unit may see status epilepticus or diabetic ketoacidosis only a handful of times a month, which turns a two-year dissertation into a race.

Then check which measurements are actually routine rather than theoretically available. Arterial blood gas analysis at defined intervals, serum lactate, daily creatinine, procalcitonin, thyroid function, formal nutritional risk scoring, structured delirium screening and bedside ultrasonography are present in some Indian units and absent in others. A protocol on delirium prevalence collapses where sedation is not scored and no validated instrument is in use, because you cannot retrofit an assessment that was never done. Where the investigation is not routine, the study becomes one that adds tests for research purposes, which changes the ethics submission and usually the cost.

Finally, check who will collect the data at the bedside and when. Critical care variables have to be captured at a fixed time point, and a resident on a rotating posting cannot personally be present for every admission. Decide early whether the design relies on nursing charts, on the unit's software, or on your own presence, because that decision shapes what the study can measure.

Cross-sectional observational and analytical designs dominate: clinical and aetiological profiles of a defined critical illness, prevalence of a complication among intensive care admissions, association between an admission parameter and a severity score, and comparison of clinical and laboratory characteristics between two defined groups. Diagnostic and prognostic accuracy studies comparing a score or bedside index against a stated outcome are equally well established, as are audit-style studies of antimicrobial concordance, transfusion appropriateness and bundle compliance.

Prospective observational designs are the norm here rather than the exception, because the time-anchored measurements these topics need are seldom recorded consistently in retrospective charts. Interventional designs comparing ventilation strategies, sedation regimens or resuscitation targets are largely outside postgraduate scope, both for sample size and for ethical reasons.

Bring three to five shortlisted titles rather than one, since guides frequently rule out a topic on grounds you could not have known — a departmental study already running, a ventilator fleet due for replacement, a change of intensivist coming that will alter unit protocols mid-study.

Settle five things in that meeting: annual admissions in the relevant category, which investigations are routine and which would have to be added for the study, the exact time point at which the primary variable will be captured and who will capture it, whether the topic requires an outcome such as mortality or intensive care stay and how that will be ascertained, and which journal the eventual paper is aimed at. Where the topic depends on another department — microbiology for culture and susceptibility, radiology for imaging severity, the blood centre for component data — secure that cooperation in writing rather than assuming informal goodwill will hold for two years.

Sometimes, but less often than in other specialities, and the reason is worth understanding before you commit. Intensive care charts are dense but they are recorded for clinical purposes, not to a research time grid. A retrospective study of lactate at a fixed point after admission usually founders because the first gas was taken at whatever moment the sample could be sent, sometimes after two hours of fluid, sometimes before any. Severity scores can rarely be reconstructed in full, because the arterial blood gas or bilirubin needed for a component was not sent on that day.

Three further cautions. Unit protocols drift. Sedation practice, empirical antibiotic policy and ventilation targets change when consultants change, so a two-year archive may contain two different practice eras; record the period per case and check for a step change before pooling. Discharge against medical advice is common in Indian intensive care units, frequently for financial reasons and frequently in the sickest patients, so any outcome-linked analysis built on records must report how many patients left before the outcome was determined and must not quietly classify them as survivors. Device-associated infection data are recorded separately from the clinical notes where an infection control team maintains surveillance, and the two sources often disagree; state which one is your data source.

Where a topic genuinely can be done retrospectively — microbiological profile and resistance patterns, indications for ventilation, aetiological spectrum of a condition — state the period searched, how cases were identified, and the completeness of retrieval.

The synopsis is the condensed two to four page document submitted for topic registration. The protocol is the full document of twelve to twenty pages containing the detailed review of literature, methodology with case definitions, measurement time points, scoring rules and instruments, statistical plan, timeline and annexures. The synopsis is normally extracted from the completed protocol.

Full institutional ethics committee approval before data collection. Four points need explicit attention in this speciality.

Capacity and surrogate consent. Most patients you enrol will be sedated, ventilated, delirious or unconscious at the moment of enrolment and cannot consent for themselves. The protocol must state who may consent on their behalf under the term your committee uses — legally acceptable representative, legally authorised representative or next of kin — define the order of precedence, and supply a surrogate consent form as a separate annexure. State also that consent is sought again from the patient personally if capacity returns during the admission, and that the patient may then withdraw their data. Committees ask about this routinely and a protocol silent on it will be returned.

The therapeutic misconception. Families consenting in an intensive care unit are under acute distress and may assume that participation improves the care their relative receives. The information sheet must state plainly, in the local language version as well, that this is an observational study, that treatment decisions rest entirely with the treating team, and that declining changes nothing about the care given.

Additional sampling. Where the study requires a blood sample beyond routine care, state the volume and justify it. Critically ill patients are already subject to substantial diagnostic phlebotomy and many are anaemic on that account, so committees scrutinise any addition. Where possible, design the study around the residual volume of samples already drawn and say so.

Data on dying and end-of-life care. Several topics touch on patients who die during the admission or on decisions to limit treatment. State how such records are handled, that no family is approached for research purposes at the time of death, and that findings are reported in aggregate.

Clearance commonly takes six to ten weeks and retrospective approval is not granted.

Because the variables move, and without an anchor the study measures nothing definite. A serum lactate taken on arrival, after a litre of fluid, and six hours into vasopressor support are three different numbers from the same patient, and a dataset that mixes all three has no interpretable relationship with anything. The same applies to the organ failure score, which is a daily measure, to the oxygenation ratio, which changes with every ventilator adjustment, and to electrolytes under active correction. Choose one convention and write it into the objectives, the methodology and the proforma: the first value within a stated window of intensive care admission, the worst value within the first twenty-four hours, or the value at the moment the case definition was met. If your primary variable is a score, state the computation window as well — the Acute Physiology and Chronic Health Evaluation II score uses the worst value of each component in the first twenty-four hours, so a patient who dies or is discharged at eight hours cannot be scored and must be handled by a pre-stated rule rather than an improvised one. Where exposure and outcome are measured at different times, the design is no longer purely cross-sectional; say so honestly in the protocol rather than letting an examiner point it out at the viva.

Substantially. A PhD proposal typically runs 6,000 to 10,000 words and carries an extended critical literature review, a theoretical framework, a detailed methodology chapter and a discussion of expected contribution to the field. An MD synopsis is a two to four page registration document. The research question can be the same; the depth expected is not.

Most universities require registration within six to nine months of joining. Shortlist in the first two months, finalise with your guide by the third, and file for ethics clearance immediately afterwards. Because most critical care topics are prospective and depend on accrual, every week of delay costs admissions you cannot recover; plan the data collection window to close at least six months before submission so that analysis and writing are not compressed into the final term.