Queen Elizabeth University Hospital: case note review - overview report

6. The Impact of Infection on Patient Outcomes

6.1 Background

Our ToR charged us with defining: a) how many children were affected by GNE bacterial infection (addressed in Chapter 4); b) whether it is possible to associate these infections with the environment of the QEUH/RHC (Chapter 5); and c) was there an impact on care and outcomes in relation to infection? This chapter addresses this third question.

The findings described in this chapter will also inform the final question asked of us: d) what recommendations should be considered by NHS GGC and, where appropriate, by NHS Scotland more generally to address the issues arising from these incidents to strengthen infection prevention and control in future? Our overall recommendations are given in Chapter 10.

The approach we took towards defining and assessing the impact of infection is described in Chapter 3, section 3.6.7. We address issues relating to aspects of clinical care in section 6.2: these are the principal items that contributed to the scoring framework we used to assess the impact of the infection. In section 6.3 we will discuss information about the 22 children known to have died by the time of the publication of this report. Our approach to the collection and grading of adverse events is described in section 6.4 and in section 6.5, we try to bring these various themes together in a narrative summary of impact.

The themes raised by families in their submission to the Panel are dealt with in Chapter 7.

6.2 Items relating to aspects of clinical care

Details of the specific items identified in this section were sought in the data collection process and included the Data Synthesis files created to inform Panel review (as described in section 3.6). The data in this section are presented in two ways, first for all episodes in the Review (correcting the numbers for those which were not evaluable); and second, comparing the episodes we considered ‘Most likely’ to have been linked to the hospital environment with the remaining episodes (as described in section 5.6).

6.2.1 Overall impact

We have described the approach taken to agree an overall impact score for each infection episode (section 3.6.7). The distribution of these scores for evaluable episodes of infection is summarised in Table 6.1 which presents the data both as the impact grade used for the panel review and as the equivalent NHS Scotland Risk Assessment Matrix score.

Table 6.1. Overall impact grade allocated in each episode of infection

Panel Impact Grade	NHSS Risk Assessment Matrix score	Whole Group (No. evaluable = 115)	Most Likely linked (No. evaluable = 36)	Least Likely linked (No. evaluable = 79)
None	1. Negligible	1 (1%)	0 (0%)	1 (1%)
Minor	2. Minor	5 (4%)	1 (3%)	4 (5%)
Significant	3. Moderate	65 (56%)	21 (58%)	44 (56%)
Severe	4. Major	40 (35%)	12 (33%)	28 (35%)
Critical	5. Extreme	4 (3%)	2 (6%)	2 (2%)
*Non evaluable		3	1	2

*Three patients were not evaluable for an overall impact grade because of the circumstances of their admission and complications of their disease.

6.2.2 Length of hospitalisation

We requested data for the duration of the whole admission during which each infection episode took place and/or was treated. We also collected details of all antibiotics used and the duration of antibiotic treatment. It became clear to us, however, that whilst the duration of the entire admission and/or the duration of antibiotic treatment was easiest to define, the best measure of the overall impact (burden) of the infection was the length of an inpatient admission that could, as far as it was possible to assess, be attributed to the treatment of the infection. Making this distinction was not always easy: in many patients, the duration of admission was extended either because of other toxicities, including other infections, or because the patient stayed in hospital to continue or restart treatment. In others, antibiotics were continued for several days (and occasionally significantly longer) after the patient had been discharged from inpatient care. However, by considering the details collected from the case notes to inform the patient’s clinical timeline, we found it was generally possible to make a reasonable assessment of the length of an admission that could be accounted for principally because of the occurrence of the GNE infection (Table 6.2).

Table 6.2 Length of hospital stay attributed to the infection

Duration	Whole Group (No. evaluable = 115)	Most Likely linked (No. evaluable = 36)	Least Likely linked (No. evaluable = 79)
1-7 days	15 (13%)	9 (25%)	6 (8%)
8-14 days	43 (37%)	11 (30%)	32 (40%)
15+ days	57 (50%)	16 (44%)	41 (52%)
Not evaluable	3	1	2

6.2.3 Removal of the central venous line

Patients with indwelling venous access devices (lines and ports) are especially susceptible to blood stream infections. It is frequently necessary to remove the device in order to eradicate a blood stream infection although there are often good clinical justifications to try to ‘salvage’ the line (or port) with antibiotic treatment in order to facilitate continuing care in a challenging clinical situation. This may be possible by extending antibiotic treatment and by using antibiotic ‘locks’ (instillation of a high concentration of an antibiotic into the catheter lumen, and allowing it to remain for a period of time), but may also be associated with risk if the strategy fails.

The removal of a central line in a child almost always requires a short anaesthetic and, under most circumstances, a replacement line will be required once the infection has been treated. This contributes a degree of further risk and an added logistical challenge to the delivery of care.

The data we collected are summarised in Table 6.3.

Table 6.3 Infection episodes requiring removal of the central line

CVL removed?	Whole Group (No. evaluable = 115)	Most Likely linked (No. evaluable = 36)	Least Likely linked (No. evaluable = 79)
Yes	78 (68%)	26 (72%)	52 (66%)
No	37 (32%)	10 (27%)	27 (34%)
No CVL in situ	2	1	1
Not evaluable	1	0	1

6.2.4 Admission for Intensive Care

Bacteraemia of any kind can result in severe illness, but many GNE bacteria can be virulent pathogens (with the potential for endotoxic shock) which may cause rapid clinical deterioration and risk of death. Admission to PICU is therefore an important measure of the severity of infection and its impact on the patient.

All infections which merit admission to PICU are serious but there are occasions when patients who might sometimes be managed satisfactorily in the normal ward environment are admitted to PICU because of the opportunity for closer and more intensive monitoring and, perhaps, short term life support. There are others whose deterioration is more profound and who may require prolonged support. Empirically, we therefore divided admissions to PICU into two groups - those of up to 3 days and those with longer stays - as a way of trying to reflect this distinction (Table 6.4). We also recognised that some patients required PICU support for other problems at the time of the bacteraemia but not, in our judgement, specifically because of the bacteraemia.

Table 6.4 Admission to PICU

PICU admission at the time of the bacteraemia	Whole Group (No. evaluable = 114)	Most Likely linked (No. evaluable = 37)	Least Likely linked (No. evaluable = 77)
Yes, for 1- 3 days	9 (8%)	6 (16%)	3 (4%)
Yes, for >3 days	3 (3%)	2 (5%)	1 (1%)
No	102 (89%)	29 (78%)	73 (95%)
Yes, unrelated to infection	3	0	3
Not evaluable	1	0	1

6.2.5 Cancer treatment disruption

Children and young people with cancer are most often treated with a predefined plan (protocol) for treatment which is shaped by the details of their diagnosis and is based on the outcome of prior experience of the same condition or by a clinical trial. These protocols represent an ‘intent to treat’ strategy which incorporate combinations of different elements of therapy (chiefly chemotherapy and/or radiation therapy and/or surgery according to diagnosis). This is delivered according to a schedule that has either been achieved in the past with defined results, or represents an ambition based on preliminary or pilot data but may still be under evaluation in a current trial. The reality, however, is that many patients are not able to adhere to the intended plan at some or other stage in their treatment, with the result that therapy has to be paused or modified, or both.

The circumstances leading to a decision to pause or modify treatment will vary but generally these relate to the extent to which the patient already manifests side effects from the therapy delivered to date. This includes, for example: the severity of bone marrow suppression with consequent low blood counts; infection; nutritional deterioration; other organ toxicity (e.g. liver or kidney function problems); and the psychological state of the patient and/or family. The oncologist treating the patient continually monitors these factors and must judge whether, and when, a pause in treatment is required, and if treatment needs to be modified in the future (for example, reduction or omission of a planned chemotherapy dose or the deferral of the start of a course of radiation therapy). This constitutes the ‘art’ as well as the ‘science’ of oncology care. Clinical trial protocols, however, usually include rules that set out whether, when and how treatment should to be modified in relation to specific toxicities.

It is to be expected that most parents and clinicians would agree that adherence to the intended treatment protocol leads to a better chance for long term disease control and cure. However, there are remarkably few peer reviewed publications that explore the impact of treatment adjustment and treatment delay on outcome. From an entirely pragmatic perspective, minor (up to one week) delays in treatment are common in practice and there is no evidence that this makes a material difference to outcome. Longer delays are, however, sometimes necessary and whilst the impact is also uncertain, this is more undesirable and would generally be avoided if circumstances permit. It is also generally the case that avoiding delay in the early phase of treatment after a new or recurrent diagnosis is more important than at later stages in treatment.

In order to explore the impact of infection on the continuity of cancer treatment, we tried to define, from the clinical records, the extent to which treatment was disrupted specifically in relation to the GNE infection. We looked principally for evidence that chemotherapy had been delayed on the basis of the infection although this was not always clear and often compounded by the fact that, for example, blood count recovery was insufficient to allow chemotherapy to proceed with safety. Complexity in attributing a causal effect is compounded when one considers that whilst infection may itself contribute to delayed bone marrow recovery, this can also happen without infection.

It was more difficult to identify where drug doses had been subsequently modified but dose reduction as a result of an infection is less likely to be required in the short term than a delay in re-starting treatment. We also recognised that not all patients were receiving chemotherapy at the time of their infection and we therefore also looked for evidence that other elements of treatment had been deferred. The data in Table 6.5 represent our best estimates of all types of treatment delay.

Table 6.5. Delay to treatment attributed to the infection

Duration of delay	Whole Group (No. evaluable = 101)	Most Likely linked (No. evaluable = 32)	Least Likely linked (No. evaluable = 69)
None	53 (53%)	18 (56%)	35 (50%)
1 - 7 days	19 (19%)	6 (19%)	13 (19%)
8 – 14 days	17 (17%)	5 (16%)	12 (17%)
15+ days	12 (12%)	3 (9%)	9 (13%)
Not evaluable*	17	5	12

*9 episodes of infection were experienced by patients with non-malignant diagnoses for whom we did not attempt to determine the impact of the infection on treatment; 3 patients were not evaluable because the data were insufficient; 3 were excluded because treatment was delayed by other toxicities and we were unable to separate the specific impact of the infection; treatment was discontinued after infection in 1 patient because of concurrent evidence of progressive disease; and in 1 patient treatment had already been completed after stem cell transplantation.

6.3 Details of the children and young people who have died

At the time of the publication of this report, we were aware of the deaths of 22 patients (6 male and 16 female) who had been included in our Review.

Dates of their death ranged from November 2016 to January 2021. The primary diagnoses were: Solid Tumour (n=7); Leukaemia (6); CNS Tumour (6); Lymphoma (2); and non-malignant condition (1).

Median age at death was 6 years 6 months with a range from 1 year 8 months to 16 years 3 months. The median interval from the last GNE infection episode to the date of death was 10 months (range 1 day to 3 years 8 months).

Three patients died within 28 days of a GNE infection episode. Two of these died from tumour related causes and their deaths were not linked to the prior infection; in one of these cases we decided that the preceding infection was Unrelated to the hospital environment and in the other that it was Probably related to the hospital environment.

The third child in this early post infection group died in PICU 6 days after the last positive culture was taken. This occurred in the very early phase of a stem cell transplant undertaken in the context of rapidly progressive disease. Although disease progression was a major factor, we judged that the GNE bacteraemia was a significant factor in the cause of death and noted that sepsis was also identified as the principal cause of death on the death certificate issued by NHS GGC. We determined that this bacteraemia infection was Probably related to the hospital environment.

One further child, whose infection we had similarly determined was Probably linked to the hospital environment, also died relatively early (within 6 weeks) of the infection episode. Death occurred in PICU 36 days after the last positive culture. There were a number of other serious contributory factors but we judged that the GNE bacteraemia was implicated in the cause of death; this was also recorded as a possible contributory factor on the death certificate issued by NHS GGC.

Overall, death certificate information was obtained for 19 of the 22 patients – it was unavailable in one because the patient had died abroad and in the other two because there was insufficient time from the notification of death to the completion of this report.

In summary, based on death certificates and clinical information, we decided that infection was implicated as a cause of death in 2 patients (discussed above) whilst; 19 had died of their underlying disease (all cancer) and 1 died from other causes unrelated to infection.

6.4 Adverse Events

The approach we took to the detection of Adverse Events (AE) by the use of the PTT and interrogation of the Datix system at NHS GGC is described in chapter 3, section 3.4.

6.4.1 PTT data

In addition to the 115^[67] GNE bacteraemias occurring in the 83 patients eligible for the PTT analysis (all of which were defined as an AE), the PTT review separately identified 386 other AE. Of these 24 (5%) were classified as Category I^[68], according to the National Framework^[69] (discussed in section 3.4.4) and occurred in 17 (14%) of the 117 episodes.

All unplanned admissions to PICU were classified as Category I AE and occurred in 16 episodes^[70], accounting for 67% of all Category I AE. Moreover, 7 of the remaining Category I events occurred in 2 of the same 16 episodes. The only Category I event to be recorded in an infection episode with no PICU admission occurred in a patient who was resuscitated for sepsis on the ward but whose condition stabilised sufficiently to avoid PICU admission. These data suggest that admission to PICU is an obvious way of identifying patients with the greatest risk of the most serious category of AE for audit and review.

There were 362 Category II^[71] AE of which 78 (22%) related to removal of the central line.

Overall, of the 501 AEs detected by the PTT, only one fifth (91 (18%)) were unrelated to management of the infections. Six of these were Category I – four of the admissions to PICU, one pulmonary embolus and one case of pressure ulcers.

We recognise that some of the triggers identified by the PTT relate to expected complications of chemotherapy or represent support measures commonly required by this group of patients. Nevertheless, the use of the PTT could provide a useful audit tool to monitor trends in the occurrence of AE that occur during care.

6.4.2 Datix system data

In total, 174 incidents were recorded in Datix in 65 (76%) of the 84 patients included in the Review (collected during the period of review) with a median of 2 (range, 1-6) incidents per patient. In 23 of these patients a total of 31 Datix reports were made during an admission that incorporated one or more episodes of Gram-negative environmental infection. The other 143 Datix reports were made during admissions that occurred either before (n=84) or after (n=59) the admissions with infection episodes.

Of the total 501 AEs detected with the PTT, only 6 (1%) were reported in Datix, which included 2 (8%) of the 24 Category I AEs. One of these patients had severe sepsis and died in PICU; this was correctly classified in Datix as Category I (Extreme). The second patient had a PICU admission for toxic megacolon due to C difficile, but this was incorrectly scored as Moderate (Category II – should have been Category I) on Datix.

The 6 AE common to both systems included 2 incidents that were categorised as infection control in Datix: septic shock associated with GNE bacteraemia and the C difficile infection (mentioned above). The other 4 incidents common to both systems were pressure ulcers, bacterial contamination of infused donor bone marrow cells, and 2 pain control incidents.

The 23 patients with Datix reports made during an admission that incorporated one or more episodes of Gram-negative environmental infection had a total of 36 incidents. However, 9 (29%) of these were recorded as Negligible risk, i.e. Category III^[72] incidents that were not associated with harm, whereas the PTT review only included Category I and II incidents. However, one of the Datix incidents that was graded as Negligible risk was one of the two deaths we identified as being associated with infection. The reason given in Datix for reporting this death was (correctly) that it had occurred within seven days of receiving donor stem cells. Whether or not the stem cell transplant per se contributed to death is not the issue we raise; it is rather that, as the incident was an unexpected death, this should have been reported as Category I.

In addition to the Extreme incident (Category I - death in PICU), there was only one other incident reported in Datix as Major in any of these patients throughout the period of the Review, and this was unrelated to an infection episode.

Of the total of 174 Datix incidents, 124 (71%) were classified as Minor or Negligible. Only 5 of the total 174 incidents were coded as relating to infection control for the entire period of the Review, and only 2 of these were documented as such during an infection episode. However, some Datix reports that were classified as ‘Other’ clearly described an infection control incident (e.g. bacterial contamination of donor stem cells) and should have been coded as such.

We concluded that Datix reporting significantly underestimated the number of AE experienced by this group of patients and that, even when reported, some incidents were incorrectly classified and under scored in terms of their severity.

6.5 Summary

In this chapter we have tried to set out measures of the burden of the GNE infections experienced by the patients we have reviewed. Our data provides an insight into the overall experience of children and young people with cancer (these were, in the great majority, children and young people with leukaemia and other forms of cancer) who experience such infections.

In the course of our review, we used selected clinical indices to express our overall assessment of the impact of an infection episode on the patient (section 3.6.7). In so doing, we identified that over one third (38%) had experienced an overall severe or critical impact and only 5% of the whole group experienced no or minor impact.

Whilst accepting the limitations on our ability to define the length of hospital admission directly attributable to infection, our estimate suggests that additional hospitalisation of 15 days or more was required in approximately half of the episodes reviewed.

Removal of a central line was required in two thirds of these episodes for the management of the infection, which implies that, in almost all those patients, a further anaesthetic and surgical procedure would have been required to insert a replacement.

Twelve patients (11%) required admission to PICU specifically for the consequences of their infection although admissions were short (1-3 days) in the majority of cases.

Finally, infection is an important reason for treatment to be disrupted in this clinical context and we estimated that approximately 30% of episodes were associated with a delay in planned treatment of over 1 week, and in 12% for over 2 weeks.

Tables 6.1 to 6.5 also analyse the data according to two groups: those with the GNE infections we determined were ‘More likely’ to have been acquired from the hospital environment (defined in section 5.6) and those for whom we did not find strong evidence for an association (labelled as ‘Less likely’).

There is, in fact, little difference between the two groups except for the frequency with which patients were admitted to PICU: 8/37 in the ‘More likely’ group vs 4/77 in the ‘Less likely’ group. This difference is significant (Relative Risk 4.16 (95% CI 1.34–12.94)) and whilst it may be unwise to speculate too much on a single variable in an analysis of this kind, variation in the type and pathogenicity of organisms contributing to these two groups may be the relevant factor. Table 5.4 shows that there was a significant excess of Stenotrophomonas spp. in the ‘More likely’ group of infection episodes. This may be relevant and, perhaps, a predictive factor for greater risk of severe illness.

GNE was implicated in the deaths of 2 of the 22 patients known to have died; this was the primary cause of death in one and an important contributory factor in the second. Both were infected with Stenotrophomonas maltophilia.

The use of the PTT identified that 5% of 501 AE identified in the whole population included in the Review were Category 1 events (classified as Major or Extreme in the NHS Scotland risk assessment matrix). Comparison of PTT data with Datix incident reporting suggests that the NHS GGC reporting system had significantly underestimated the true extent of such events and, where reported, may underestimate their severity.

Finally, we recognise that nothing analysed in this chapter measures the broader implications of infection on the lives of the children and young people affected, and their families. Unplanned or prolonged admission, or both, will contribute to the already significant impact they experience in their lives. It further disrupts schooling, social life, parental work, and the care of siblings or dependent relatives. It contributes to additional anxiety both because families are well aware that infection is a risk, can be serious and may be life threatening; and because families are anxious about the consequences of delays to treatment.

We have been able to characterise part of the physical impact of infection but wish to emphasise that the emotional, social, financial and psychological costs can also be significant.